The iPDU Handbook
A Guide to Intelligent
Rack Power Distribution
The iPDU Handbook
A Guide to Intelligent
Rack Power Distribution
Raritan began developing KVM switches for IT professionals to manage servers remotely in 1985. Today, as a brand
of Legrand, we are a leading provider of intelligent rack PDUs. Our solutions increase the reliability and intelligence
of data centers in 9 of the top 10 Fortune 500 technology companies. Learn more at Raritan.com
©2016 Raritan Inc. All rights reserved. Raritan® is a registered trademarks of Raritan Inc. or its wholly-owned
subsidiaries. All others are registered trademarks or trademarks of their respective owners.
V1195R2
Table of Contents
1.0 Introduction
..................................................................................................................................... Page 5
............................................................................................. Page 9
2.0 Fundamentals and Principles
............................................................................Page 11
2.1 Overview and Class of Devices
2.2 Electrical Terminology ..................................................................................................Page 13
2.3 Electrical Power Distribution to the Rack .......................................................Page 14
.........................................................................................Page 17
2.4 Plugs, Outlets, and Cords
.........................................................................................................Page
21
2.5 Ratings and Safety
................................................................................................... Page 23
2.6 Overload Protection
.......................................................................................................Page 27
3.0 Elements of the System
...........................................................................................................................
Page 29
3.1 Rack PDU
.................................................................................Page 37
3.2 Environmental Management
.....................................................................................................Page 40
3.3 System Connectivity
.............................................................................................Page 41
3.4 Rack PDU Management
.............................Page 47
4.0 Considerations for Planning and Selecting Rack PDUs
4.1 Power Available and Distributed to Racks .................................................... Page 49
4.2 Power Requirements of Equipment at Rack .................................................Page 53
......................................................................................................Page 55
4.3 Rack PDU Selection
.............................................................................................................Page 58
4.4 Power Efficiency
............................................................................................Page 63
5.0 Future Trends for Rack PDUs
5.1 Higher Density, Higher Power Rack PDUs with Sensors ...................... Page 66
5.2 Increased Intelligence at the Rack
to Support Efficiency Initiatives ............................................................................... Page 67
5.3 Integration with Higher Level
Data Center Management Systems .....................................................................Page 68
....................................................................................................................................................Page 70
Reference
..................................................................................................................Page 73
Key Terms to Know
References and Further Reading .....................................................................................Page 77
................................................................................................................................................Page 78
Index
1.0 Introduction
2.0 Fundamentals and Principles
5
The iPDU Handbook
T
he rack power distribution unit (rack PDU) has emerged from obscurity. As the
last link of the elaborate data center power chain, the traditional role of the rack
PDU has been to deliver stable, reliable and adequate power to all the devices
in the rack or cabinet—servers, storage, network equipment—which are plugged
into it. Although it provides the electrical heartbeat to all the systems that run the critical
applications that support the operation of the business (or that, in some cases, are the
business); it was often considered a simple commodity, just a power strip. Typically, IT
merely told facilities how much power was needed, based on device nameplate specs—and
often with redundancy, so there was plenty of headroom and minimal risk of downtime.
Little thought was given to efficiency or what other value a rack PDU could provide. That
was yesterday.
Over the past few years, system availability has become a “given” and data center
management attention is now being focused on operational costs, efficiency improvements
and resource optimization. With the annual expenditure for powering the average data center
surpassing the cost to purchase the IT equipment itself, the use (and waste) of energy is now
targeted as a priority. Beyond the actual cost to power the data center there are the related
issues that impact both current operations and future expansion—e.g. physical space and
utility power availability, CO2 footprint and potential government regulation. Since almost all
of the power delivered from the utility to the data center is consumed either directly by the
devices plugged into rack PDUs or indirectly by the infrastructure to bring power to the rack
and cool the devices, the once obscure rack PDUs have become visible on the data center
management radar.
Not surprisingly, many of the major strategies to address the above issues and improve
overall data center efficiency depend on new capabilities not available in the commodity
outlet strips of a few years ago.
To consider a few of these capabilities:
ƒƒ In order to maximize the use of data center space and other resources there has been
a trend to deploy racks densely packed with 1U servers or power-hungry blade servers.
Today’s rack PDUs typically handle loads of 5-10 kW with 20 outlets and there are PDUs
now designed to support 20+ kW and 36 or more outlets.
ƒƒ To increase IT staff productivity and conserve energy by employing lights-out and/or
remote data center operation, some rack PDUs provide real-time monitoring, reporting
and alerts as well as secure, reliable outlet switching.
ƒƒ To identify ghost (no definable function), underutilized or grossly inefficient servers for
possible elimination, replacement, consolidation or virtualization, some rack PDUs
provide individual outlet monitoring.
ƒƒ To create individual awareness, accountability and/or chargeback for power usage and
CO2 footprint, some rack PDUs are equipped with highly accurate, real-time power
measurement capabilities at the PDU and outlet level.
ƒƒ To optimize IT workload and make informed decisions for infrastructure capacity
planning, IT and facilities managers need rack PDU management software that
continually collects data on power consumption, analyzes trends and correlates with IT
workload data.
Introduction
7
Why Has the Selection of Rack PDUs Become So Important?
As you can imagine, a wide variety of rack PDU configurations is available based on
parameters such as: number of phases, voltage, total amps, branch circuits, number of
outlets, socket type, plug type, rack units consumed, and physical dimensions. Beyond the
functions of the basic rack PDU, additional capabilities are available in rack PDU categories,
we call metered, switched, and intelligent. Furthermore, if you cannot find an off-the-shelf
rack PDU that matches your specific requirement, some vendors will assemble or even
design a custom rack PDU (also called BTO/ETO: built-to-order/engineered-to-order).
In the following sections we will discuss the basic concepts, considerations and approaches
in designing, selecting and deploying the appropriate rack PDU for typical data center
applications. We will describe not only the basic ingredients for delivering adequate, reliable
power to the rack, but also the factors and best practices that will contribute to a reliable,
operationally efficient and environmentally sound data center for today and for the future.
8
The iPDU Handbook
2.0 Fundamentals and Principles
2.0 Fundamentals and Principles
9
2.0 Fundamentals and Principles
2.1 Overview and Class of Devices
2.2 Electrical Terminology
2.3 Electrical Power Distribution to the Rack
2.4 Plugs, Outlets, and Cords
2.5 Ratings and Safety
2.6 Overload Protection
10
The iPDU Handbook
IT
equipment is normally mounted in racks or cabinets with provisions for all
necessary cables, ventilation, cooling and convenient access. There are large
data center PDUs that are used earlier in the power chain and take the form
of panel boards mounted on walls or free standing pedestals. In this handbook
we’re discussing only the rack PDU, at the end of the chain, which supplies power to the IT
equipment in the rack. Unless otherwise stated any reference to “PDU” for the remainder of
this handbook means “rack PDU.”
Rack PDUs come in many configurations with respect to number and type of receptacles,
voltage, load capacity and physical mounting (horizontal or vertical). A unit may perform
no function other than providing power to the devices plugged into it; or it may also provide
additional functions—for example, turning power off and on remotely, monitoring power
consumption and sensing the temperature in an IT equipment rack.
2.1 Overview and Class of Devices
A rack PDU is mounted in an IT equipment rack and provides electrical power to various IT
devices such as servers, networking, and storage equipment. Today, rack PDUs are available
in a number of configurations. We describe below the basic characteristics of four types of
rack PDUs.
2.1.1 Types of Rack PDUs
Rack PDUs can be divided into two categories: Non-Intelligent PDUs and Intelligent PDUs
Non-Intelligent PDUs
ƒƒ Basic PDUs: Are power strips that are used in critical environments such as data
centers. They distribute correct voltage and current to multiple outlets to power IT
equipment in racks.
ƒƒ Monitored PDUs: Allow a user to view a local display that typically provides electric
current information only. However, this information cannot be accessed remotely as the
units have no network connectivity capabilities.
Intelligent PDUs
ƒƒ Metered Input PDUs: Meter power at the PDU-level and can display the data both locally
and over a network. Metering helps users to determine power usage and available
capacity at the rack and facilitates provisioning. By metering at the input level, users
can avoid overloading circuits and more easily calculate efficiency metrics like Power
Usage Effectiveness (PUE).
ƒƒ Metered Outlet PDUs: Meter power at the outlet-level and can display the data both
locally and over a network. Like metered input PDUs, outlet-metered models help
users to determine power usage and available capacity at the rack, and also facilitate
provisioning. Most importantly, outlet-level metering allows users to understand power
consumption at the device or server-level which makes it possible to allocate costs to
specific business units or customers.
2.0 Fundamentals and Principles
11
ƒƒ Switched PDUs: Offer the features of Metered Input PDUs and also provide controlled
on/off switching of individual outlets. They enable authorized users to securely power
cycle devices remotely in a specific order, offer power sequencing delay to minimize
inrush currents, prevent unauthorized device provisioning, and can power off devices
that are not in use to conserve energy.
ƒƒ Switched PDUs with Outlet Metering: Combine all of the capabilities of Switched PDUs
with those of Outlet -Metered PDUs.
Non-Intelligent PDUs
Core Features
Basic
Monitored
Intelligent PDUs
Metered
Input
Metered
Output
Switched
Switched
and Outlet
Metered
Power Distribution
Input Metering
Outlet Metering
Network Connectivity
Switching
Secondary Features
Environment Sensor
Support
Strong Passwords
Encryption
User Permissions
Additional Ports, e.g. USB
Always
Sometimes
Never
Figure 1: Types of Rack PDUs
(Source: Raritan, Inc.)
To learn more about intelligent PDUs, visit: www.raritan.com/ipdus
12
The iPDU Handbook
2.2 Electrical Terminology
It will be helpful to review several electrical term definitions before discussing rack PDUs
in more detail.
Voltage (Volt): Electromotive force, or difference in electrical potential, measured in volts
and equal to the current times the resistance.
Current (Amp): The flow or rate of flow of electrons, ions, or holes in a conductor or medium
between two points having a difference in potential, measured in amperes and equal to the
ratio of the voltage to the resistance.
Active power (Watt): The real power drawn by an IT device. This determines the actual
power purchased from the utility company and the heat loading generated by the equipment
since, for IT equipment, 1W of electricity equals 1W of heat.
Apparent power (VA): The product of the voltage applied to the IT device times the current drawn
by the equipment. The VA rating is used for sizing wiring and circuit breakers. The apparent
power is always equal to or larger than the active power.
Power factor: The ratio of active power to apparent power.
Energy measurement (kW): Electrical energy is measured at one instant in time in watts
or kilowatts. For example, a 100W light bulb consumes 100 watts at any instant in time, but
energy is consumed, and billed for by utilities, over time. A kilowatt-hour (kWh) is a unit of
electrical energy or work, equal to the power supplied by one kilowatt for one hour, e.g. a
1000W light bulb left on for one hour, or a 100W light bulb left on for ten hours.
Line: An electrical conductor which is a source of voltage, e.g. 120V. In a single-phase
system there are one or two lines. In a three-phase system there are three lines.
Neutral: An electrical conductor that provides a return path for the voltage supplied by a
line. The neutral itself is not a source of voltage.
Ground: A conducting body, such as the earth or an object connected with the earth, whose
potential is taken as zero and to which an electric circuit can be connected. The purpose of
a ground wire is to safely direct stray currents to ground rather than allowing them to pass
through someone contacting the stray currents.
4-wire and 5-wire systems: A 4-wire rack PDU consists of one ground wire and three lines
(see Three phase Delta below), each line carrying equal voltage but each voltage sine wave
is 120 degrees out of phase with the others. The voltage of two lines is available (line to line,
e.g., L1-L2). A 5-wire system is the same as the 4-wire system but with the addition of a
neutral wire (see Three phase Wye below) so that the voltage of one line can be supplied (line
to neutral) as well as the voltage of two lines (line to line).
Three-phase Delta “∆“: This configuration gets the name Delta because a schematic
drawing of it has three transformers forming a triangle or the Greek letter Delta. The three
lines connect to the three “corners” of the triangle.
Three-phase Wye “Y“: This configuration gets the name Wye because a schematic drawing
of it has three transformers meeting in the center forming the letter “Y”. The three lines
connect to the three “branches” of the “Y” and the neutral connects to the center.
2.0 Fundamentals and Principles
13
2.3 Electrical Power Distribution to the Rack
2.3.1 Branch Circuits
Power is distributed to the rack over one or more electrical branch circuits. Branch circuits
are power feeds that originate from a panel, switch or distribution board and terminate into
an electrical receptacle mounted in a junction box near the IT equipment rack. Depending
on the data center’s layout, branch circuit wiring can be overhead, underneath a raised floor
or both. The rack PDU itself could have multiple branch circuits. See Section 2.6 for details
regarding branch circuit protection requirements.
2.3.2 Branch Circuit Load Capacity
The power that can be delivered by a branch circuit depends on the electrical characteristics
of the circuit. A key factor in delivering power to a rack is whether the power is single phase
or three phase. The amount of electricity delivered to a rack is often referred to as the load
capacity and is the product of the rated voltage and the rated current and is presented as
Volt-Amps (VA) or kVA (VA x 1000). Given the rated voltage and current, the load capacity that
can be delivered by a branch circuit is determined using these formulas:
ƒƒ Single phase: Load Capacity = Rated Voltage x Rated Current
ƒƒ Three phase: Load Capacity = √3 x Rated Voltage x Rated Current
2.3.3 Branch Circuits: Rated Voltage
The rated voltage of a branch circuit specifies both its magnitude (volts) and number of
phase conductors. Single-phase wiring is straight forward and consists of two wires (plus
safety ground) where the AC voltage is a single sinusoidal wave as measured across the
two wires.
Three-phase wiring is more complicated and consists of either three (three-phase
conductors) or four (three-phase and one neutral) wires, plus safety ground. Three-phase
branch circuits deliver more power, but require a rack PDU specially designed for threephase branch circuits. Internally, a three-phase rack PDU divides the 3 or 4 branch circuit
wires into pairs of single phase circuits – and these single-phase circuits are wired to the
rack PDU’s single-phase outlet receptacles.
The three-phase conductors have the same voltage magnitude but the sinusoidal AC
waveforms are out of phase with each other by 120 degrees. Regardless of the number of
wires, the rated voltage of three-phase wiring is always the measured voltage difference
between any two-phase conductor wires – not the difference between a phase wire and
neutral. Just as with single-phase power described above, connecting across one 120V hot
line and the neutral provides 120V AC. Connecting across any two 120V hot lines, say L1 and
L2, provides 208V AC, not 240V AC. Why? Because the phase of L1 is offset 120 degrees from
L2 the voltage is not 240V (120V x 2), as it is for single-phase, but is 120V x √3 or 120V x 1.732
= 208V. A three-phase PDU can deliver three circuits of 208V each. Some rack PDUs take
advantage of a neutral wire to provide three circuits of both 120V and 208V. But as mentioned
in the preceding paragraph, regardless of the number of wires, or whether or not both a
higher and lower voltage are supplied as outputs, a three-phase rack PDU is rated at the
voltage between two phases, e.g., L1 and L2 which in the example here is 208V.
14
The iPDU Handbook
A rack PDU can also provide 400V AC. As with the 208V three-phase rack PDU, if one of
those lines is connected to a neutral instead of another line, it provides a single-phase
output circuit-- which for a 400V-rated PDU is 230V AC (400V / 1.732 = 230V). This is a
common deployment in Europe and is becoming more common for high-power racks in
North America.
Figure 2: Three-Phase Wiring Diagram
(Source: Raritan, Inc.)
When looking at three-phase rack PDU specifications you will often see the terms Wye
and Delta or the Greek letters Y and Δ. These terms or letters were chosen because the
electrical configuration diagram of a Delta transformer looks like a Δ and the electrical
configuration diagram of a Wye transformer looks like a Y. A rack PDU which does not
convert a higher input voltage, e.g., 208V or 400V, to a lower output voltage, e.g., 120V or
230V, but instead retains the higher voltage throughout the PDU uses a Delta transformer.
A Delta transformer has three connection points, one at each corner of the triangle. Each
of these points is a connection for one of the three lines. Connecting any point to any
other point provides a line-to-line connection, e.g., L1 to L2 and provides 208V or 400V as
described in the examples above.
A rack PDU which does convert a higher input voltage to a lower output voltage uses a Wye
transformer. A Wye transformer has three connection points for the lines, one at the end of
each “arm” of the Y and one at the “foot” of the Y. The center intersection point of the Y is a
fourth connection point and is where the neutral wire is attached. Connecting any two of the
three line connections together, e.g., L1 and L2, provides 208V or 400V. Connecting any one
of the three line connections to the neutral, e.g., L1 and neutral, provides 120V or 230V as
described in the examples.
2.0 Fundamentals and Principles
15
Rated Voltage
Location
Number of Wires
Outlet Voltage(s)
120V
North America
3 (line + neutral + ground)
120V
208V
North America
3 (line + line + ground)
208V
230V
International
3 (line + neutral + ground)
230V
208V 3Ø ∆
North America
4 (3 lines + ground)
208V
208V 3Ø Y
North America
5 (3 lines + neutral + ground)
Mixed 120V & 208V
400V 3Ø Y
International &
North America
5 (3 lines + neutral + ground)
230V
Figure 3: Branch Circuit Rated Voltage and Wire Requirements
(Source: Raritan, Inc.)
2.3.4 Branch Circuits: Rated Current
The amount of current that can flow in a circuit is determined by the size (thickness) of
its wire and terminating receptacle. All branch circuits are required to be over current
protected using a circuit breaker (or fuse). The rating of the circuit breaker is sized to the
current carrying capacity of the branch circuit’s wiring and receptacle. For example, 10 AWG
(American Wire Gauge) wire and a NEMA L21-30R receptacle are both specified at 30A – so
a branch circuit using these components must be protected by a 30A circuit breaker.
In North America, the national electric code for data centers (NEC Article 645) requires
branch circuit wiring to be rated 125% greater than the total connected load. To ensure this
requirement is met without having to run heavier gauge wires, all electrical devices (rack
PDUs, computers, etc.) used in North American data centers must be certified to UL 609501. UL 60950-1 limits a device to draw no more than 80% of the rating of its input plug. For
example, a rack PDU (or any other device) containing a 30A NEMA L21-30P plug must not
draw more than 24A. This 80% limitation is commonly known as “derated” current.
16
The iPDU Handbook
The following table summarizes power available for various branch circuits.
Location
Rated
Voltage
Nominal Current
Rated Current
120V
North America
208V
1.9kW
20A
16A
208V 3Ø
North America
3.3kW
5.8kW
230V
International
Available Power/
Branch Circuit
3.7kW
16A
16A
400V 3Ø
11.1kW
120V
2.9kW
208V
30A
24A
208V 3Ø
8.6kW
230V
International
5.0kW
7.4kW
32A
400V 3Ø
32A
22.2kW
Figure 4: Branch Circuit Available Power
(Source: Raritan, Inc.)
2.4 Plugs, Outlets, and Cords
Rack PDUs are available with several types of plugs and receptacles (or outlets), designed
so that only the appropriate rack PDU plug will fit into the appropriate circuit outlet and only
the appropriate device plug will fit into the appropriate rack PDU receptacle. This is done to
protect equipment, e.g. so that a device that is designed for 120 volts only isn’t accidentally
plugged into a 208-volts circuit, and for safety reasons; e.g. so that a server that draws 30
amps doesn’t overload a circuit designed to handle only a maximum of 15 amps.
The two major classifications of plugs and receptacles used in data centers are defined
by NEMA (National Electrical Manufacturers Association) and IEC (International
Electrotechnical Commission). NEMA plugs and receptacles are most common in North
America and IEC plugs and receptacles are most common in Europe. However, many data
centers in North America use IEC plugs and receptacles and there are many families of
plugs and receptacles in use in data centers around the world.
A significant concern in data center power distribution is unintentional disruption of power
by accidental disconnecting of cords. A variety of solutions exist that lock the plug into the
receptacle and prevent the cord separating from the receptacle.
2.0 Fundamentals and Principles
17
There are three methods of securing the plug in the receptacle.
1.Plug with tabs snaps into the receptacle locking them together. One example of this
system is Raritan’s SecureLockTM.
2.Plug inserted into a receptacle with a locking mechanism that grips the plug ground
blade.
3.Wire retention clips mounted to the PDU chassis hold the plug in the receptacle.
The higher the current carrying capability of a plug, receptacle, or cord, the greater the
amount of wire conducting material, typically copper, is required to prevent overheating the
wire that could lead to an electrical fire. Note that the smaller the wire gauge number, the
greater the diameter of the conductor.
The conductors are surrounded by insulating material and jacket, which may have special
properties. For example, the jacketing may be designed to resist damage from exposure to
oil. Typical insulating and jacket materials are PVC, rubber, and neoprene.
The number of wires in a cable can vary. Below are some typical data center configurations:
ƒƒ Two wires: One hot and one neutral wire without a ground wire.
ƒƒ Three wires: One hot, one neutral, and one ground wire.
ƒƒ Four wires: Three hot wires (L1, L2, L3) and one ground wire.
ƒƒ Five wires: Three hot wires (L1, L2, L3), one neutral wire, and one ground wire.
18
The iPDU Handbook
RECEPTACLE
IEC 60320, C-13
IEC 60320, C-19
PLUG
IEC 60320, C-14
IEC 60320, C-20
IEC 60320, C-15
IEC 60320, C-14
IEC 60320, C-15
IEC 60320, C-16
RATING
15 Ampere 250 Volt
UL/CSA
10 Ampere 250 Volt
International
20 Ampere 250 Volt
UL/CSA
16 Ampere 250 Volt
International
15 Ampere 250 Volt
UL/CSA
10 Ampere 250 Volt
International
RECEPTACLE
IEC 60320, C-5
IEC 60320, C-7
PLUG
IEC 60320, C-6
IEC 60320, C-8
IEC 60320, C-13
IEC 60320, C-18
IEC 60320, C-17
IEC 60320, C-18
IEC 60309
4H-R
IEC 60309
4H-P
15 Ampere 250 Volt
UL/CSA
10 Ampere 250 Volt
International
30 Ampere 250 Volt
UL/CSA
32 Ampere 230 Volt
European “CE” Mark,
VDE
IEC 60309
6H-R
IEC 60309
6H-P
2.5 Ampere 250 Volt
UL/CSA
2.5 Ampere 250 Volt
International
IEC 60309
4H-P
20 Ampere 250 Volt
UL/CSA
16 Ampere 230 Volt
European “CE” Mark,
VDE
IEC 60309
6H-R
2.5 Ampere 250 Volt
UL/CSA
2.5 Ampere 250 Volt
International
30 Ampere 125 Volt
UL/CSA
20 Ampere 125 Volt
UL/CSA
IEC 60309
4H-R
RATING
IEC 60309
6H-P
Figure 5: IEC Plugs and Receptacles
(Source: Raritan, Inc.)
RECEPTACLE
PLUG
RATING
RECEPTACLE
PLUG
NEMA 5-15R
NEMA 5-15P
NEMA L6-15R
NEMA L6-15P
20 Ampere 125 Volt
USA & Canada
Locking receptacle and plug.
Polarized.
(UL 498)
15 Ampere 250 Volt
USA & Canada
Receptacle and plug.
Polarized.
(UL 498)
NEMA 6-15R
NEMA L5-20R
NEMA 6-15P
NEMA L5-20P
20 Ampere 250 Volt
USA & Canada
Locking receptacle and plug.
Polarized.
(UL 498)
20 Ampere 125 Volt
USA Receptacle and plug.
Canada - Plug only.
Polarized.
(UL 498)
NEMA 5-20R
NEMA L6-20R
NEMA 5-20P
NEMA L6-20P
30 Ampere 125 Volt
USA & Canada
Locking receptacle and plug.
Polarized.
(UL 498)
20 Ampere 250 Volt
USA Receptacle and plug.
Canada - Plug only.
Polarized.
(UL 498)
NEMA 6-20R
NEMA L5-30R
NEMA 6-20P
NEMA L5-30P
30 Ampere 250 Volt
USA & Canada
Locking receptacle and plug.
Polarized.
(UL 498)
15 Ampere 125 Volt
USA & Canada
Locking receptacle and plug.
Polarized.
(UL 498)
NEMA L5-15R
NEMA L5-15P
RATING
15 Ampere 250 Volt
USA & Canada
Locking receptacle and plug.
Polarized
(UL 498)
15 Ampere 125 Volt
USA & Canada
Receptacle and plug.
Polarized
(UL 498)
NEMA L6-30R
NEMA L6-30P
Figure 6: NEMA Plugs and Receptacles
(Source: Raritan, Inc.)
2.0 Fundamentals and Principles
19
TYPE SOOW - 600 VOLT - UL/CSA
CATALOG
NUMBER
NO. OF
COND.
AWG
SIZE
COND.
STRAND
02763
2
18
02769
3
18
02770
4
02722
02765
NOM. INS.
THICKNESS
NOMINAL O.D.
CURRENT
AMPS
APPROX. NET
WEIGHT LBS/M
STD.
CTN.
8.76
10
70
250’
9.27
10
80
250’
0.390
9.91
7
95
250’
0.76
0.370
9.40
13
80
250’
0.76
0.390
9.91
13
95
250’
0.030
0.76
0.420
10.67
10
115
250’
0.045
1.14
0.510
12.96
18
135
250’
250’
INCHES
mm
INCHES
mm
16/30
0.030
0.76
0.345
16/30
0.030
0.76
0.365
18
16/30
0.030
0.76
2
16
26/30
0.030
3
16
26/30
0.030
02766
4
16
26/30
02723
2
14
41/30
02762
3
14
41/30
0.045
1.14
0.535
13.59
18
170
02768
4
14
41/30
0.045
1.14
0.575
14.61
15
205
250’
02724
2
12
65/30
0.045
1.14
0.570
14.48
25
195
250’
250’
02725
3
12
65/30
0.045
1.14
0.595
15.11
25
225
02726
4
12
65/30
0.045
1.14
0.650
16.51
20
270
250’
02767
2
10
104/30
0.045
1.14
0.620
15.75
30
250
250’
250’
02728
3
10
104/30
0.045
1.14
0.660
16.76
30
290
02727
4
10
104/30
0.045
1.14
0.715
18.16
25
355
250’
16063
3
8
133/29
0.060
1.52
0.855
21.72
40
485
250’
250’
16064
4
8
133/29
0.060
1.52
0.980
24.89
35
670
16065
5
8
133/29
0.060
1.52
1.075
27.30
28
790
250’
16073
3
6
133/27
0.060
1.52
0.980
24.89
55
700
250’
16074
4
6
133/27
0.060
1.52
1.080
27.43
45
875
250’
16075
5
6
133/27
0.060
1.52
1.200
30.48
36
1015
250’
16083
3
4
133/25
0.060
1.52
1.140
28.96
70
920
250’
16084
4
4
133/25
0.060
1.52
1.260
32.00
60
1150
250’
250’
16085
5
4
133/25
0.060
1.52
1.365
34.67
48
1400
16093
3
2
133/23
0.060
1.52
1.330
33.78
95
1355
250’
16094
4
2
133/23
0.060
1.52
1.460
37.08
80
1690
250’
16095
5
2
133/23
0.060
1.52
1.580
40.13
64
1960
250’
Figure 7: Cord Specifications
(Source: General Cable)
20
The iPDU Handbook
2.5 Ratings and Safety
Rack PDUs, like all other electrical equipment, are subject to many general and specific
safety standards. There are general industry terms and conventions that should be
understood in order to ensure a reliable and safe data center. These are discussed in detail
below.
2.5.1 Nameplate Data
Nameplate data is the electrical power consumption information specified by the equipment
manufacturer. It is typically a conservative estimate of the maximum amount of power the
device could draw. This information is found on a label near the electrical power input to the
IBM Server Model 520 - Rack Mounted Drawer
Condition
Configuration
Typical Heat
Release
(Voltage 110V)
Airflow
Nominal
Overall System Dimensions
(W x D x H)
Weight
Maximum
at 35°C
watts
cfm
(m³/hr)
cfm
(m³/hr)
lbs
kg
in
mm
Minimum
420
25
44
40
68
117
53
25 x 37 x 23
630 x 939 x 584
Full
600
30
51
45
76
117
53
25 x 37 x 23
630 x 939 x 584
Typical
450
25
44
40
68
117
53
25 x 37 x 23
630 x 939 x 584
Airflow Diagram Rackmount
Cooling Scheme F-R
Configuration
Description
Model
ASHRAE
Class
Minimum
1-way, 1.5 GHz processor
15 GB memory
3
Full
2-way, 1.65 GHz processor
maximum memory
Typical
1-way, 1.65 GHz processor
15 GB memory
Front to Rear
F-R
device. There will be more discussion of the use of nameplate data below.
Figure 8: Nameplate Data
(Source: IBM)
2.5.2 Power Rating vs. Load Capacity
There can be confusion about power capacities and load capacity. The confusion stems from
a misunderstanding of approval agency regulations and from some manufacturers who
may use misleading terminology. In North America typical circuits have a maximum current
carrying capability, and use circuit breakers or fuses rated at 15A, 20A, 30A, etc. In other
words, a 20A fuse will blow or a 20A circuit breaker will trip if a 20A circuit experiences more
than 20A for some period of time. The period depends on the magnitude of the current and
the type of fuse or circuit breaker protecting the circuit.
2.0 Fundamentals and Principles
21
In North America, circuits are to be loaded to 80% of their maximum capacity. For example,
a 15A circuit should not carry more than 12A, a 20A circuit not more than 16A, a 30A circuit
not more than 24A, etc. The 80% value, e.g., 16A for a 20A circuit, is often referred to as the
derated value or the load capacity. In North America a rack PDU vendor’s specifications
sheet may have a few current carrying specifications.
The specifications provided and the terminology used may vary by vendor but the following
is a typical example:
ƒƒ Maximum line current per phase: 30A
ƒƒ Rated current: 24A (30A derated to 80%)
ƒƒ Maximum current draw: 6 x 16A (six circuits, each capable of carrying up to 16A)
In Europe and other parts of the world circuits are simply described at their rated capacity,
e.g., 16A and 32A.
As mentioned above, apparent power is specified in volt-amps (VA) which is volts x amps.
Load capacity is specified in VA where the amps are the rated current, i.e., the derated
value. For example, for a single-phase rack PDU with a nominal voltage of 208V and a rated
current (not the maximum current) of 24A, the load capacity is 5.0 kVA (208V x 24A).
2.5.3 Approval Agencies
In order to meet applicable local and national electric codes, rack PDUs must be safe
and not emit electromagnetic radiation. Recognized approval agencies are contracted by
manufacturers to test products according to existing standards. A product that passes
agency testing receives an approval listing number, and the manufacturer can then affix
the agency approval listing logo on each product. The listing logo is your assurance that
the product meets applicable safety and electric codes. The manufacturer is required, upon
request, to provide you the listing number and a copy of the testing report. You can also
submit the listing number to the approval agency to verify compliance.
2.5.4 Proper Grounding
The National Electric Code (NEC Article 645.15) requires all exposed non-current-carrying
metal parts of an information technology system to be grounded. This means all equipment
within a rack and the metal rack itself must be grounded.
The inlet plug of a PDU contains a ground pin. When this plug is connected to a properly
wired receptacle, the PDU becomes the grounding point for the equipment plugged into the
rack PDU. The PDU can also be used to ground the metal rack and most PDUs contain a
special threaded hole for this purpose. Typically, a grounding wire is connected to the rack
and the PDU using screws. Care should be taken to make sure any paint on the rack is
scraped off where the grounding wire is attached to ensure proper electrical conduction.
There are also special grounding screws with teeth under the head of the screw to ensure
a good ground connection.
22
The iPDU Handbook
Approval
Description
Standard/Revision/Year
Comment
UL
Safety
UL 60950-1
Required in USA
cUL/CSA
Safety
CAN/CSA-C22.2 No. 60950-1-03
Required for Canada
CB
Safety
IEC 60950-1
Common replacement for
UL, CSA & CE in countries
that accept CB
CE
Electromagnetic Interference (EMC)
EN 5502:2006
Europe
CE
Safety
EN 60950-1
Europe
FCC-A or B
Electromagnetic Interference (EMC)
FCC 47 CFR Part 15
USA
ICES-003
Electromagnetic Interference (EMC)
ICES-0003 issue-004
Canada
Figure 9: Safety and Electromagnetic Approval Agencies
(Source: Raritan, Inc.)
2.6 Overload Protection
The Underwriters Laboratories standard UL 60950-1 applies to the safety of information
technology equipment (ITE) and requires the use of branch circuit over-current protection
for ITE PDUs greater than 20 amps. Typically, ITE PDUs greater than 20 amps and certified
after April 2003 must have built-in UL 489 circuit breakers or fuses (e.g., UL 248-5 fuses)
suitable for branch circuit protection.
UL 60950-1 permits products at a maximum current of 15 and 20 amps without circuit
breakers or fuses, since the 15 or 20 amp circuit breakers in the building are considered
sufficient to protect the PDU; however, supplementary protection in the PDU provides
additional protection. UL also “grandfathers” PDUs at more than 20 amps that were
certified prior to April 2003. Although such PDUs are still being sold, their use should be
avoided if they are to be incorporated in larger ITE systems designed to the latest UL 609501 standard.
Newly certified ITE PDUs at more than 20 amps are required to use over-current protection
that meets branch circuit protection requirements in accordance with the National Electrical
Code, ANSI/NFPA 70. In effect, this means these products are required to have branch
circuit breakers listed under UL 489, “Standard for Molded-Case Circuit Breakers, MoldedCase Switches and Circuit Breaker Enclosures” or fuses listed for branch circuit protection,
such as those listed to UL 248-5, “Low-Voltage Fuses – Part 5: Class G Fuses.”
In addition to standard UL 489, Underwriters Laboratories also publishes the standard UL
1077, “Standard for Supplementary Protectors for Use in Electrical Equipment.” Devices
certified to this standard are called “Supplementary Protectors” and are called “Recognized”
2.0 Fundamentals and Principles
23
components, not “Listed” devices, as are UL 489 breakers. UL Listed Circuit Breakers meet
more stringent requirements for branch circuit protection than Supplementary Protectors
with UL Recognition.
Circuit breakers are used in a variety of ways. They are mounted in panel boards (also
referred to as building PDUs) and rack PDUs to protect branch circuit wiring. They are
also built into equipment to protect components and systems. Interrupting a short circuit
– current flow limited only by the resistance of wiring – is a severe test of a circuit breaker.
If the interrupting capacity of the breaker is not adequate, the device can literally explode.
UL 489 requires the breaker to be functional after being subjected to a short-circuit test.
UL 1077 and the IEC standard EN 60934 allow for breakers to clear a short-circuit condition
but become safely destroyed in the process. UL 489 breakers can interrupt short circuits of
5,000 amps or more. Typically, UL 1077 breakers can interrupt fault currents of 1,000 amps.
Overloads can be short term or long term. The protective device must not trip with a
momentary or short-term over-current event that is normal for the piece of equipment being
protected. Servers, for example, may create inrush currents as their internal power supply
and filter circuits start. These inrush currents typically last only a fraction of a second and
seldom cause a problem. If an overload lasts longer than a few minutes, the breaker should
open to prevent overheating and damage. What gives a breaker the ability to discriminate
between normal and damaging over-currents is its delay curve.
If you would like more information on the topic of data center overload protection see, “Data
Center Power Overload Protection: Circuit Breakers and Branch Circuit Protection for Data
Centers” at www.raritan.com/high-power-white-paper.
24
The iPDU Handbook
Chapter 2 Summary
Rack PDUs come in many configurations with respect to number and type of receptacles,
voltage, load capacity and physical mounting.
ƒƒ Rack PDU can be divided into two categories:
Non-Intelligent PDUs
Basic PDUs - Are power strips that are used in critical environments such as data
centers. They distribute correct voltage and current to multiple outlets to power IT
equipment in racks.
Monitored PDUs - Allow a user to view a local display that typically provides electric
current only. However, this information cannot be accessed remotely as the units
have no network connectivity capabilities.
Intelligent PDUs
Metered Input PDUs - Meter power at the PDU-level, and can display the data
both locally and over a network. Metering helps users to determine power usage
and available capacity at the rack, and facilitates provisioning. By metering at the
input-level, users can avoid overloading circuits and more easily calculate efficiency
metrics like Power Usage Effectiveness (PUE).
Metered Outlet PDUs - Meter power at the outlet-level, and can display the data both
locally and over a network. Like metered input PDUs, outlet-metered models help
users to determine power usage and available capacity at the rack, and facilitates
provisioning. Most importantly, outlet-level metering allows users to understand
power consumption at the device or server-level which make it possible to allocate
costs to specific business units or customers.
Switched PDUs - Offer the features of Metered Input PDUs and also provide
controlled on/off switching of individual outlets. They enable authorized users to
securely power cycle devices remotely in a specific order, offer power sequencing
delay to minimize inrush currents, prevent unauthorized device provisioning, and
can power off devices that are not in use to conserve energy.
Switched PDUs with Outlet Metering - Combine all of the capabilities of Switched
PDUs with those of Outlet-Metered PDUs.
ƒƒ Power is distributed to the rack over one or more electrical branch circuits.
ƒƒ Single-phase wiring is straight forward and consists of two wires where the AC voltage
is a single sinusoidal wave as measured across the two wires.
ƒƒ Three-phase wiring is more complicated and consists of either three or four wires, plus
safety ground. Many PDUs support both single-phase and three-phase power.
ƒƒ When looking at three-phase rack PDU specifications, a Delta transformer looks like a Δ
and the electrical configuration diagram of a Wye transformer looks like a Y.
2.0 Fundamentals and Principles
25
ƒƒ The amount of current that can flow in a circuit is determined by the size (thickness) of
its wire and terminating receptacle.
ƒƒ Rack PDUs are designed so that only the appropriate rack PDU plug will fit into the
appropriate feeder circuit outlet and only the appropriate device plug will fit into the
appropriate rack PDU receptacle.
ƒƒ In North America circuits are to be loaded to 80% of their maximum capacity. For
example, a 15A circuit should not carry more than 12A. This limited value is known as
the “derated” current.
ƒƒ The National Electric Code (NEC) requires all exposed non-current-carrying metal parts
of an information technology system to be grounded.
ƒƒ Underwriters Laboratories requires the use of branch circuit over-current protection for
PDUs greater than 20 amps.
ƒƒ Overloads can be short term or long term. The protective device must not trip with a momentary
or short-term over-current event that is normal for the piece of equipment being protected.
26
The iPDU Handbook
3.0 Elements of the System
3.0 Elements of the System
27
3.0 Elements of the System
3.1 Rack PDU
3.2 Environmental Management
3.3 System Connectivity
3.4 Rack PDU Management
28
The iPDU Handbook
R
ack PDUs are the final endpoint of power supplied to IT equipment from incoming
building feeds through a chain of equipment including UPS, transformers, and
larger PDUs and circuit panels. IT and facilities management are increasingly
viewing their rack PDUs not merely as a collection of power outlets for IT
equipment but as a network of critical devices that significantly impact the overall efficiency
and effectiveness of the data center. As such they need to be properly managed like the IT
equipment they power. This is driving the trend for use of more intelligent PDUs in data
centers with environmental sensors and even integration with higher level data center
management systems. This section describes not only the components of the physical rack
PDU and basic environmental sensors, but also the rack PDU management system that
leverages the intelligence in PDUs for operational improvements and energy use reduction.
Further, this system can interface with and become part of a larger ecosystem of enterprise
IT and facilities management systems.
3.1 Rack PDU
Over the past few years, average power consumption per server has rapidly increased
with the adoption of high-power computing equipment like blade servers and data center
containers. In addition, ongoing deployment of densely packed storage, virtualization, and
cloud computing results in data centers with greater watts per sq. ft. requirements from
more densely packed racks such as a rack filled with 1U servers. To support new, powerhungry IT equipment, data center managers have to deliver more power to the IT equipment
rack. Over the last decade, the typical power required at a rack has increased from 2
kilowatts to 12 kilowatts and continues upward.
3.1.1 Single-Phase or Three-Phase Input Power for Rack PDUs
To accommodate the increased power demands at IT equipment racks, data center
managers are deploying rack PDUs capable of supplying multiple circuits, higher voltages
and higher currents. One way to increase the power at the rack is to increase the number of
circuits and the voltage coming to the rack.
The amount of power available for use is referred to as apparent power and is calculated as
volts x amps and is described as volt-amps or VA. A 120V, 20A circuit has an apparent power
of 2400VA or 2.4kVA. A 208V, 20A circuit has an apparent power of 4160VA or 4.2kVA. Thus
one 208V circuit provides almost twice as much power as one 120V circuit assuming the
current (amperage) remains the same. With three 208V circuits it is clear that a substantial
amount of power can be deployed in one three-phase PDU.
The cable to provide power to a three-phase PDU is thick and heavy but not as thick and
heavy as the multiple, individual cables required to provide the same amount of apparent
power using either single-phase 120V or single-phase 208V. Running a single three-phase
power cable to each three-phase rack PDU reduces both the number of cables, making
installations easier, and the physical bulk of the cables, so less space is filled with cables
that can block necessary cooling airflow under raised floors and within IT equipment racks.
In cases where power needs to be provided at 120V for devices such as routers, hubs and
switches, as well as at 208V for demanding servers, three-phase PDUs can provide outlets
with both 120V (one of the three lines and a neutral) and 208V (two of the three lines). Threephase power at the IT equipment rack is a convenient way for data center managers to
efficiently deploy both greater power capacity and flexibility.
3.0 Elements of the System
29
3.1.2 Form Factor
Rack PDUs are available in heights of one rack unit (1U; 1.75 inches) or two rack units (2U;
3.5 inches) for horizontal mounting in a 19-inch equipment rack.
Zero U rack PDUs mount vertically, typically to the vertical rails at the back of the rack.
This can offer advantages. Zero U PDUs don’t consume any rack unit spaces, and since the
receptacles on the Zero U PDU line up better with the power cords for each IT device in the
rack, they allow for the use of shorter power cords. This results in neater cable arrangements
contributing to better airflow within the rack, which can improve cooling efficiency.
Depending on the rack cabinets, Zero U rack PDUs can be mounted with screws or hung into
the cabinet via buttons, that are spaced 12.25 inches apart.
Higher-power rack PDUs will commonly be equipped with circuit breakers for branch
circuit protection. These circuit breakers may cause the rack PDUs to extend deeper into
the racks. Considerations should be made for how these PDUs are mounted in the rack and
whether outlets are facing center or back to allow for cable management, airflow, and easy
accessibility and serviceability of the IT equipment.
3.1.3 Outlet Density and Types
Rack PDUs vary in the number of outlets supported based on the physical size (length, width
and depth) and the total space available for mounting outlets and internal components and
the power handling capacity of the PDU. For example, a 1U rack-mount PDU may have
enough space for eight 120V/15A NEMA 5-15R outlets; whereas a 2U rack-mount PDU may
have enough space for twenty 120V/15A NEMA 5-15R outlets. On the other hand, a Zero U
PDU may have 24 IEC C-13 230V/10A outlets or just four 250V/30A NEMA L15-30R outlets
to support blade servers.
In the case of a large number of devices, each demanding a moderate amount of power,
a large number of moderate power outlets are required. A typical dense “pizza box”
deployment would include two rack PDUs for redundant power where each PDU is loaded
to 40% so that if one power feed fails, the other feed will not exceed the NEC requirement
of 80% (for North America). Typical outlets for “pizza box” servers are IEC C-13 (up to 250V,
16A) and NEMA 5-20R (up to 125V, 20A, 16A rated).
In the case of high power consumption at a rack for a few devices, each of which consumes a
lot of power (blade servers, storage, or network devices), the total amount of power required
might be comparable to the high outlet density example above, but the number and type of
outlets may be different. Density for devices such as blade servers depends on their number
of power supplies (often between two and six for redundancy), how the power supplies are
configured (power supplies are most efficient when they operate close to their maximum
level), and how many devices will be deployed in a rack.
In the case of a few devices demanding a lot of power, a large number of outlets may not be
needed but outlets capable of delivering substantial power may be required. Typical outlets
for high-demand devices, such as blade servers at 208V or 230V, are IEC C-13 (up to 250V,
16A) or C-19 (up to 250V, 32A) or, less commonly, NEMA L6-20R (up to 250V, 20A, 16A rated)
or L6-30R (up to 250V, 30A, 24A rated) locking outlets.
30
The iPDU Handbook
3.1.4 Connectors: Ethernet, Serial, Sensor, USB and Other
Today, only the very basic rack PDUs have no external connectors. Most now include a
variety of connectors based on application requirements. Below we describe four rack PDU
connector configurations and general applications.
ƒƒ No connectors: Cannot be managed externally and may not feature a local display.
ƒƒ Local navigation: Provides users with a navigable display via local buttons. User can see
basic data at the PDU or outlet-level.
ƒƒ Serial RS232 Connector: Used for local metering; may be an LCD or LED monitor. Can
be plugged into a terminal or console server for Telnet or SSH remote access. Access
via a menu or command line interface using terminal emulation. Local buttons allow
navigation to see basic unit data. No SNMP support available for alarms, unless via a
specially developed serial console server. Typically non-switched.
ƒƒ Ethernet (RJ-45) and RS232 Serial (DB-9M) Connectors: For remote metering for
PDUs, circuit breakers, and outlets. USB-A (host) and USB-B (device) connectors to
support PDU-to-PDU cascading, webcams and wireless networking. SNMP support
available for alarms, Telnet or SSH access possible for command line access. Support
for environmental sensors like temperature, humidity, airflow, air pressure and others
may be available on the PDU or with an add-on external device. Remote metered models
typically have an LCD or LED display with buttons for navigation to see basic unit and
outlet data.
3.1.5 Branch Circuit Protection
Since April 2003, Underwriters Laboratories (UL) has required branch circuit protection,
either circuit breakers or fuses, for PDUs where the inlet current is greater than the outlet
current, e.g., 30A (24A rated) plug, 20A (16A rated) outlets. 15A and 20A (12A and 16A rated)
rack PDUs can be supplied without branch circuit breakers because circuit breakers in
upstream panel boards are deemed to provide the necessary protection. Rack PDUs with
breakers or fuses are like mini-subpanels. For example, a 208V 30A (24A rated) 3-phase
PDU has 3 circuits and each circuit / set of outlets has a 20A circuit breaker.
There are four types of circuit breakers--thermal, magnetic, thermal-magnetic, and
hydraulic-magnetic.
ƒƒ Thermal circuit breakers incorporate a heat-responsive bimetal strip or disk. This
technology has a slower characteristic curve that discriminates between safe, temporary
surges and prolonged overloads.
ƒƒ Magnetic circuit breakers operate via a solenoid and trip nearly instantly as soon as
the threshold current has been reached. This type of delay curve is not ideal for servers
since they typically have inrush currents anywhere from 30 percent to 200 percent above
their normal current draw.
ƒƒ Thermal-magnetic circuit breakers combine the benefits of thermal and magnetic
circuit breakers. These devices have a delay to avoid nuisance tripping caused by normal
inrush current, and a solenoid actuator for fast response at higher currents. Both
thermal and thermal-magnetic circuit breakers are sensitive to ambient temperature.
A magnetic circuit breaker can be combined with a hydraulic delay to make it tolerant
of current surges.
3.0 Elements of the System
31
ƒƒ Hydraulic-magnetic breakers have a two-step response curve. They provide a delay on
normal over currents, but trip quickly on short circuits and are not affected by ambient
temperature.
Circuit breakers used in rack PDUs are typically thermal-magnetic or hydraulic-magnetic
with delay curves that allow for reasonable inrush currents while protecting devices from
excessive fault currents.
Int= 1.13 x In for t >1h
Int= 1.45 x In for t >1h
� 2.55 x In :t = 1– 60 s(In < 32 A)
t = 1– 120 s(In < 32 A)
� Type B: 3 x In : t > 0.1 s
� Type B: 5 x In : t > 0.1 s
� Type C: 5 x In : t > 0.1 s
� Type C: 10 x In : t > 0.1 s
� Type D: 10 x In : t > 0.1 s
� Type D: 20 x In : t > 0.1 s
Figure 10: Trip Delay Curve
(Source: Moeller Electric)
Fuses are also acceptable for PDU circuit protection. However, replacing a fuse can be time
consuming and may require an electrician leading to longer mean time to repair (MTTR).
Spare fuses must be stocked in inventory and the correct fuse must be used to ensure
reliability and protection.
32
The iPDU Handbook
The following are some points to consider when selecting a rack PDU:
ƒƒ Compliance with the latest fuse and circuit breaker standards
ƒƒ The acceptable MTTR for fuse replacement vs. circuit breaker resetting
ƒƒ Impact on uptime service level agreements if a fuse blows vs. a circuit breaker trip.
3.1.6 Circuit Breakers: Single Pole vs. Double and Triple Pole
An important consideration is the reliability and flexibility of the branch circuit breaker
configuration. Typically, circuit breakers are available as single, double or triple-pole devices.
Single-pole breakers are appropriate for circuits comprised of a hot wire and neutral, e.g.,
120V at 20A or 230V at 16A. Single-pole breakers provide a disconnect for the single hot
wire used in circuits with a hot wire and neutral. Double-pole breakers provide a disconnect
for circuits comprised of two hot wires, e.g., 208V at 20A. Some PDU designs use doublepole (or triple-pole) breakers to provide protection for 2 different circuits, e.g. two different
hot wires. Since one double-pole breaker is less expensive than two single-pole breakers,
this will lower the cost of the design. Double-pole breakers will trip if either of the two
circuits they protect are overloaded. It is less expensive than 2 (or 3) single-pole breakers
but, unless the poles can be operated independently, in a maintenance shutdown or trip, all
2 or 3 circuits are de-energized.
For example, assume a rack PDU with six branch circuits is protected by circuit breakers.
Some rack PDUs in this configuration may protect the six circuits with three double-pole
circuit breakers – one double-pole circuit breaker for the circuits with Line 1, one for the
circuits with Line 2 and one for the circuits with Line 3. It is less expensive to use doublepole circuit breakers but there are some drawbacks. As noted, double-pole breakers will trip
if either of the two circuits they protect are overloaded. This means double-pole breakers
are less reliable. Double-pole breakers are also limiting because if for instance you choose
to shut off a circuit for maintenance, you have no choice but to shut off both circuits.
Alternatively, some rack PDUs protect the six circuits with six single-pole circuit breakers –
one breaker per circuit. This is more expensive but single-pole breakers are more reliable
and less limiting. Thus, consider rack PDUs that allow only one circuit to be de-energized
for improved reliability and flexibility.
3.1.7 Circuit Breaker Metering
Circuit breaker metering is a useful feature on any rack PDU. It is particularly important
when dealing with high power because the consequences of tripping a breaker can be
disastrous if it means losing several blade servers. With circuit breaker metering the end
user sets a threshold. When that threshold is crossed, an alert is delivered so the end user
knows power demand needs to be reduced or there is the risk of tripping a circuit breaker.
Monitoring branch circuit breakers is important since high power draw means a greater
chance of tripping a breaker.
Line metering, intended for three-phase rack PDUs, is very useful for balancing the power
drawn over each line. Overdrawing power from one line relative to another line wastes
available power and unbalanced lines can place excessive demands on the neutral in
Wye-configured PDUs.
3.0 Elements of the System
33
3.1.8 Cord Length and Feed
Rack PDU power cords vary in length depending on the whips (power cables from a building
PDU) and the location of the racks. The rack PDU power cord must be long enough to reach
its power source, which is typically a whip located under the raised floor or an outlet just
above the rack. A common power cord length is ten feet (3m), but other lengths can often be
specified to a UL maximum of 15 feet (4.5m).
Rack PDU power cords may exit the PDU itself from the rear, the front, the top, or the
bottom. With the power cable exiting the bottom of a Zero-U PDU, the data center manager
will need to ensure sufficient space for the bend radius of the cable. In general, a bend
radius of 5.25 inches (3U) will be sufficient, but this should be confirmed as bend radii will
depend on the gauge (AWG) of the cord. A smaller bend radius may be acceptable for thin
cables and a larger bend radius may be required for heavy-duty cables. The orientation of
the PDU power cord may seem trivial but it can be a potential problem depending on the
physical rack and the location of the power source for the rack. Consider the orientation of
the power cord and how it will be routed to connect to the whip. For example, does the power
come up from the raised floor or down from cable trays above the racks and is there room
inside the rack to route the cable so as not to block airflow?
3.1.9 Cord Retention
Proper PDU cord retention practices, just like rack cable management, can make a big
difference in operational efficiency and reliability. Taking steps to support, organize and
secure the many power cords using some method of cord retention will dramatically improve
your ability to access and manage the equipment connected to PDUs inside the rack. This
will also minimize the chance of inadvertently unplugging power cords from rack PDUs. You
should neatly arrange and secure the power cords between equipment and the rack PDU to
allow for maximum airflow.
There are a few different methods to secure IT equipment power cords. Two of the most
common are to design the rack PDU with locking IEC C-13 and C-19 outlets that work
on standard IEC power cords. Another option, which is less expensive, is to use specially
designed outlets and power cords with tabs that securely hold the cord in place until the tabs
are released. (See Figure 11) These cords can be purchased in different colors to help identify
which power supply is plugged into the PDU.
34
The iPDU Handbook
Figure 11: SecureLockTM
(Source: Raritan, Inc.)
3.1.10 Local Display and User Interface
Virtually all rack PDUs designed for data center use have built in displays, typically LEDs,
to show current draw for the entire PDU unit. Local displays have limited functionality
compared to the information and control available from a remote interface, but they can
be convenient and useful when working at the rack itself. The local display might allow an
IT admin to toggle between current draw and voltage; or, for those rack PDUs that monitor
individual outlets, to sequence through the outlets to determine the current being drawn by
each IT device. Some switched and intelligent models will have LED indicators next to each
outlet to display status, whether it is on/off, booting, firmware upgrade, or fault.
In addition to a local display on the rack PDU itself, some PDUs offer a serial interface for
local terminal connectivity via a laptop for configuration, diagnostics or connectivity to a
serial console server that concentrates multiple connections.
3.1.11 Remote User Interface
For a remotely accessible rack PDU (all but the basic PDU or PDU with metering and only
local display) there are typically two choices for a remote user interface to the rack PDU over
an IP network. The most common is a web-based graphical user interface to an Ethernetenabled PDU. Some PDUs support SSL encrypted access (using https), while others support
only unencrypted access (using http). Check your organization’s security requirements when
selecting a PDU.
The PDU can also be accessed via Ethernet over IP using SSH (encrypted) or Telnet
(unencrypted) with a Command Line Interface (CLI). Security considerations should be kept
in mind before enabling/disabling Telnet access. Some PDU manufacturers provide a serial
console server that connects to the PDU locally via serial (RS232) and allows access to the
unit remotely using SNMP or CLI.
3.0 Elements of the System
35
One factor to consider is integration with central directory services for user authentication
and access control. This becomes especially important when the rack PDU offers the ability
to remotely turn on/off or recycle individual outlets or groups of outlets. Remote access to
the PDU does not eliminate the need for some local access to the PDU with an LED/LCD
and associated buttons.
Figure 12: Remote User Interface
(Source: Raritan, Inc.)
3.1.12 Inline Meters
For data centers with existing PDUs that lack any metering capability, adding in-line meters
can be useful to monitor power consumption per line or per circuit. In-line meters are
usually one line or circuit in and one line or circuit out. Some vendors offer models that
support up to four in’s and four out’s. There are basic inline meters that provide simple
current metering, with or without IP connectivity. Others are more sophisticated and provide
richer data like active power, apparent power, and kWh metering; and some might even
include integrated environmental monitoring. By upgrading older, basic PDUs with in-line
meters with IP connectivity, they can be managed along with metered and intelligent rack
PDUs by a rack PDU management system so that you have a comprehensive view of the
health and usage of power for day-to-day operations and planning.
36
The iPDU Handbook
3.2 Environmental Management
With the IT industry’s increasing focus on improving data center efficiency, more rack PDU
manufacturers are offering environmental sensors. These include sensors to measure
rack air temperature at the server inlets, humidity, airflow, vibration, smoke, water, and air
pressure. Some PDUs will have pre-installed sensors; others provide for optional, plug-in
external sensors. Another common approach is to deploy a completely independent rack
management system, choosing from a wide range of environmental sensors; however, this
has the disadvantage of consuming additional rack space for the rack management system
as well as the cost of a separate infrastructure—e.g. IP addresses, Ethernet ports, and
cabling. Connectivity for sensors is typically either via RS485 or 1-Wire®.
3.2.1 Temperature Sensors
Temperature sensors monitor the air inlet temperatures at IT devices such as servers (See
the sensor placement diagram on the next page). Since IT equipment generates considerable
heat, manufacturers specify a range of acceptable temperatures for proper operation. A
sensor-capable PDU should allow thresholds to be set for sending automatic alerts when
the inlet temperature approaches the vendor-specified maximum to prevent servers from
shutting down or failing due to overheating. In addition, it is also a good practice to set
a minimum temperature threshold to provide alerts when the inlet temperature is colder
than necessary. From a data center plant perspective, the cost of cooling and moving air
is the largest infrastructure expense, so maintaining IT inlet air temperatures colder than
necessary merely wastes energy and money. Temperature sensors at the rack also provide
early warning about temperature extremes, hotspots or cold spots, and can help identify
when an HVAC system is becoming unbalanced. To ensure that IT equipment is getting
enough cool air, the American Society of Heating, Refrigerating, and Air-Conditioning
Engineers (ASHRAE) has recommended that temperature probes be placed at specific
locations at the inlets of equipment racks.
3.0 Elements of the System
37
Front to Rear
(F-R)
Front to Top
(F-T)
Front to Top and Rear
(F-T/R)
Intake
HVAC
Discharge
Rack with 2 Servers
Rack with 19 Servers
Figure 13: Recommended Temperature Sensor Placement
(Source: ASHRAE)
38
The iPDU Handbook
3.2.2 Humidity Sensors
Classes (a)
Understanding the basics of what humidity is and how it affects your server room can impact
how long your computer equipment lasts and how much your electricity bill costs. Humidity
is a measurement of moisture in the air. High humidity can cause condensation buildup on
computer components, increasing the risk of an electric short. Likewise, if the humidity
is too low, data centers can experience electrostatic discharge (ESD). Humidity can be
monitored per area or zone to ensure that it is in the safe range. ASHRAE has recommended
ranges for the data center that should be consulted. Appropriate thresholds and alarms
should be set to indicate a potential problem.
Equipment Environmental Specifications
Product Operations (b)(c)
Dry-Bulb
Temperature
(°C) (e) (g)
Humidity Range,
non-Condensing
(h) (i)
Product Power Off (c) (d)
Maximum Dew
Maximum
Point (°C)
Elevation (m)
Maximum
Rate of
Change (°C/
hr) (f)
Dry-Bulb
Temperature
(°C)
Relative
Humidity (%)
Maximum Dew
Point (°C)
Recommended (Applies to all A classes; individual data centers can choose to expand this range based upon their analysis
described in this document)
A1 to
A4
18 to 27
5.5ºC DP to 60% RH
and 15ºC DP
A1
15 to 32
20% to 80% RH
17
3050
5/20
5 to 45
8 to 80
27
A2
10 to 35
20% to 80% RH
21
3050
5/20
5 to 45
8 to 80
27
A3
5 to 40
-12°C DP & 8% RH
to 85% RH
24
3050
5/20
5 to 45
8 to 85
27
A4
5 to 45
-12°C DP & 8% RH
to 90% RH
24
3050
5/20
5 to 45
8 to 90
27
Allowable
B
5 to 35
8% RH to 80% RH
28
3050
NA
5 to 45
8 to 80
29
C
5 to 40
8% RH to 80% RH
28
3050
NA
5 to 45
8 to 80
29
Figure 14: Recommended and Allowable Humidity and Temperature Ranges
(Source: ASHRAE)
3.2.3 Airflow Sensors
Airflow sensors will detect a reduction of air movement that might create the potential for
overheating, which can destroy IT equipment. There are two primary areas for monitoring
airflow in the data center–above the floor (monitored at a number of points), and below the
floor (monitored at select points). Differential airflow sensors are used to ensure that the
pressure differential between the subfloor and the floor is sufficient to control air flowing
from the subfloor to the floor above. Blockages in under-floor supply plenums can cause
high pressure drops and uneven flow, resulting in cold spots in areas where cooling air is
short circuiting to the return path. Airflow sensors should have thresholds set, and alarms
enabled, like other environmental sensors, to ensure that data center managers are alerted
when conditions are less than optimal for efficient cooling.
3.0 Elements of the System
39
3.2.4 Air Pressure Sensors
It is important to have the appropriate air pressure in under-floor supply plenums, but
sometimes this is treated as an afterthought. Air pressure that is too high will result in both
higher fan costs and greater leakage that can short circuit cooling air, while pressure that
is too low can result in hot spots at the areas most distant from the cool-air supply point.
This can lead to poor efficiency ‘fixes’ to correct the problem, such as a lowering the supply
air temperature or overcooling the full space just to address a few hot spots. Differential air
pressure sensors can be used to ensure that the pressure differential between the subfloor
and the floor is sufficient. Maintaining appropriate room pressure prevents airborne
particulates from entering the data center.
3.2.5 Contact Closure Sensors
Contact closure sensors can be used for a variety of applications. For example, a contact
closure could send an alert when a cabinet door is opened and trigger a webcam to take a
picture. Contact closure sensors can be connected to any device that can open or close a
contact--making them very versatile.
3.2.6 Other Sensors
There are a variety of other sensors that can be used in the data center. Examples include
in-cabinet smoke, water, and vibration sensors. Like the other sensors mentioned above,
these are used to send alarms when measured conditions are outside the range for proper
data center operation.
3.3 System Connectivity
3.3.1 Physical Topology
Like many functions of data center management, best practices for the physical topology
and, by extension, remote management of rack PDUs, is evolving. The current best practice
is to connect all remotely accessible rack PDUs to the “management network” (separate
from the “production network”) directly in order to collect periodic meter readings, get
immediate notifications of any faults or potential problems and enable remote power cycling
of IT equipment (depending on the intelligence of the rack PDU). When planning for a new
facility it is good practice to provide for a minimum of two Ethernet drops for each cabinet or
rack, since each will typically require two PDUs.
However, some data centers try to reduce the number of Ethernet drops and IP addresses to
minimize cabling and costs. In these cases the data center deploys PDUs that can be daisy
chained or cascaded so that one Ethernet connection can communicate with all the PDUs
in the chain.
3.3.2 Communication Protocols
The communications protocols used are typically TCP/IP when PDUs are Ethernet
connected, and proprietary protocols for PDUs serially connected to a console server, which,
in turn, connects to the TCP/IP network via Ethernet. Most often, SNMP protocol is used for
management, while LDAP and Active Directory are used for authentication, authorization,
40
The iPDU Handbook
and access control. SSH and Telnet may be used for command line management and HTTP/
HTTPS for web-based access, in some cases with SSL encryption protocols.
There are rack PDUs now with USB-A (host) and USB-B (device) ports that can be used to
support devices such as webcams and Wi-Fi modems. Some rack PDUs support MODBUS,
a common, older building management communication protocol, and some rack PDUs
support the GSM modem protocol so that cell phones can receive one-way text alerts.
3.3.3 Managing the Rack PDU
As mentioned earlier, the management system for data center power is often run on a
“management network” separate from the production network. This reduces the likelihood
of a Denial of Service (DOS) or other attack that would affect this critical function. In mission
critical facilities, there are often two connections to each rack PDU equipped with remote
communications: one for syslogging, SNMP traps, access via Web browser, and kWh logging,
and another for critical functions like remote power cycling, status of circuit breakers,
and load monitoring (normally at the unit or circuit level). In some cases, administrative
functions, like rack PDU configuration, are performed via command line scripting through a
secondary interface such as a serial port, while Ethernet remains the primary interface for
all other functions.
Some important management functions are listed below:
1. Audit logging: To track activity like switching of outlets, configuration changes, etc. Two
or more syslog servers are often used for this function.
2. Fault management: Done via SNMP with tools like HP Openview, IBM Tivoli, and others.
SNMP V2 is still the most commonly used, but SNMP V3, with its built-in security, is
recommended for applications requiring outlet control.
3.Configuration: Setup via Web browser, SNMP, command line, or a central software tool.
4. Firmware upgrades: Not an issue for older PDUs with minimal functionality, but
something that may be required for Ethernet-enabled PDUs. A central tool is essential
to manage large numbers of PDUs and to simplify management and reduce cost of
ownership.
5.Alerts: Normally via SMTP messages.
A combination of some or all of the above capabilities is required to effectively manage a data
center. Check your application requirements and choose the PDU type appropriate for your
application. If you have multiple rack PDUs (40+) you will want to consider a comprehensive
rack management system, as discussed in the following section.
3.4 Rack PDU Management System
A rack PDU management system is a software application (sometimes delivered as a software
appliance) that consolidates all communications with your rack PDUs and in-line meters
equipped for remote communication. Its main functions are data collection, reporting, power
control, element management and fault management. The system collects and converts
raw power data into useful information and provides a central point for secure access and
control across multiple rack PDUs with operation validation and an audit log. Consequently,
it simplifies the management of rack PDUs and alerts you to potential incidents. We include
3.0 Elements of the System
41
it in this section because for larger data centers with more than 40 racks it is a “must have”
in order to realize several of the benefits offered by metered, switched or intelligent PDUs-improved energy efficiency, increased uptime and lower operational costs.
Figure 15: Rack PDU Management Topology
(Source: Raritan, Inc.)
3.4.1 Data Collection
Data collection is the fundamental component that enables all reporting and most other
management functions. Of course, the management system can collect only the data
elements provided by the managed PDUs. As discussed earlier, basic PDUs provide no data,
metered PDUs provide total unit data and intelligent PDUs may provide individual outlet
data and more, so it is important to understand what data you will want to analyze when
selecting rack PDUs.
Typical data elements you will probably want to collect, as available from the PDUs, include:
ƒƒ Total unit active and apparent power.
ƒƒ Line current and capacity.
ƒƒ Outlet-level current and active power.
ƒƒ Environmental sensor data.
ƒƒ Real-time kWh metering data.
42
The iPDU Handbook
Next you will want to determine the granularity of the data you need. Your management
system should offer a user-configurable data polling interval. For most applications, a
normal polling interval is 5 minutes, which means the system will collect data points every
five minutes, but if you require greater granularity, your rack PDU may need to be able to
store data readings so that the network is not overloaded with polling traffic. You will want
to use a roll-up algorithm to collect data for long periods without causing the database to
balloon and effect performance. For most energy management applications, actual data
readings are rolled up to a maximum, average and minimum, on an hourly, daily and monthly
basis.
Advanced polling options enable a customer to minimize network traffic while still enabling
granular data collection. Advanced polling requires a rack PDU that has the memory
capacity to record and store readings called samples. The rack PDU management system
should offer the ability to configure optional sample rates for each rack PDU and also set
optional polling intervals for the management system itself to collect the stored samples at
each rack PDU not previously collected. For example a rack PDU can be configured to record
and store samples every minute. The rack PDU management system can be configured to
poll the rack PDU once an hour. In each poll, it will pull the 60 one-minute samples since
the last poll, with the intelligence to know the last reading it recorded on the previous poll.
3.4.2 Reporting and Analytics for Power Monitoring and Measurement
Reporting and graphing should include active power, current, temperature, humidity, and
information derived from the basic collected data--such as energy usage, cost and carbon
emitted due to the energy consumed--for standard and selected time periods.
Reports on maximums and minimums for current, temperature, humidity and active power
simplify key tasks and ensure that you are not in danger of exceeding circuit breaker ratings,
overcooling or undercooling. For instance, environmental information can give data center
operators the confidence to raise temperature set points without the risk of undercooling
IT equipment. Analyzing trend line graphs, reports, and “what-if” modeling can help you
perform capacity planning based on real world data.
Outlet-level data and reporting granularity can help you become more energy efficient. It
enables you to determine the potential savings of upgrading to more energy efficient servers
or the benefits of server virtualization. Consolidating several low-utilization physical servers
as virtual servers on one high-utilization physical server can reduce overall expenses, but
you will need to understand the resulting power demand of the host servers. You can also
establish objectives, report on usage and implement changes for both physical components
of the data center--floor, room, row, rack and IT device; and, also logical groupings like
customer, department, application, organization, and device type. This level of detail creates
visibility and accountability for energy usage. Some IT organizations issue energy bill-back
reports to users/owners of the IT equipment.
3.4.3 Graphical User Interface (GUI)
The graphical user interface is your window into all of the rack PDU management system
functions. This should be clean, intuitive, and web-based, functioning with all major web
browsers. A web-based system provides you more remote access options and is easier
3.0 Elements of the System
43
to support and upgrade. The GUI will most likely include a user-configurable dashboard.
The dashboard can be displayed in the data center network operations center for an easy,
at-a-glance view of the status of the data center power and environmental conditions.
This will give your customers, internal and external, a good indication of your data center
management capabilities.
Figure 16: Rack PDU Management System Graphical Interface
(Source: Raritan, Inc.)
3.4.4 Element Management
The main components of element management include centralized rack PDU access and
control, firmware management, and bulk configuration. You can view all your managed rack
PDUs from one web browser window and get a summary view of the name, location, status,
manufacturer model, and firmware level. You will want to be able to drill down to manage
at the PDU unit level, line level and, in many cases, outlet level. One click, sign-on access to
each managed PDU can give you control through the PDU’s own GUI.
Since intelligent PDUs run firmware with many configuration options, the rack PDU
management system should allow you to centrally store rack PDU firmware/configuration
versions and facilitate distribution to multiple PDUs. Configuration template storage
and distribution will simplify initial PDU installation as well as future unit additions and
replacements.
3.4.5 Fault Management
Rack PDU management systems often provide both a map view and a floor layout view, and
use a color scheme to provide an at-a-glance view of the health of all managed PDUs. Health
problems are discovered in several ways. The system can receive an SNMP trap or a syslog
44
The iPDU Handbook
event so that you are alerted of the problem as it happens. A management system can also
poll the rack PDU at set intervals to collect the heath status of the communication path to the
PDU or critical failures, and forward events to a higher-level enterprise management system.
3.4.6 Local and Remote Control/Switching
Switched rack PDUs allow for outlet control including on/off power cycling. However, most
IT devices have more than one power supply for power redundancy purposes and these
supplies are connected to outlets on separate Rack PDUs. Through the management
system, you can power cycle at an IT device level, which will programmatically switch outlets
from multiple PDUs. The system should also allow for grouping of IT devices into racks so
you can control a full rack. Finally, any switched PDU must allow for flexible sequencing and
delay so that an in-rush current spike does not trip a circuit breaker and so that application
intra-system dependencies are taken into account during startup and shutdown.
3.4.7 Security of Data and User Access
Remote monitoring, metering and management require secure remote access via Ethernet
and/or serial connections. To ensure security an intelligent rack PDU should have strong
encryption and passwords and advanced authorization options including permissions,
LDAP/S and Active Directory. A web session timeout will protect against a user accidentally
leaving an authenticated session live while not in use.
3.4.8 Administration and Maintenance
Most of the administration takes place during the initial set-up. All systems will allow for
GUI entry of this data, but that can be time consuming. Systems should also allow for the
import of configuration information; e.g. via CSV files. During the set-up you will add your
rack PDUs and the hierarchical and logical relationships. Hierarchical relationships include
data center, floors, rooms, rows, racks, rack PDUs and IT devices. Logical associations
include owners/customers of the IT device and IT device type. The administrator will also
set the data pruning intervals to ensure unnecessary data is pruned from the system.
3.4.9 Open Point of Integration
Most data centers have some other management systems already in use, so it is important
that the rack PDU management system can be integrated into these systems to minimize
the amount of duplicate data entry and collection. Asset management and enterprise
reporting systems are two typical systems that should logically interface with the rack PDU
management system. The asset management system will automatically add rack PDUs, IT
devices and their associated connections to the rack PDU management system’s inventory
of managed devices. Integration with an enterprise reporting system enables the creation
of custom reports with the additional ability to correlate data that exits in other systems.
In recent years, a more comprehensive class of products called data center infrastructure
management (DCIM) have been introduced that normally include the above functions,
overall capacity planning tools and more.
Learn more about DCIM at www.RaritanDCIM.com
3.0 Elements of the System
45
Chapter 3 Summary
ƒƒ Rack PDUs are the final endpoint of power supplied to IT equipment from incoming
building feeds. As a result, IT data centers increasingly view rack PDUs as an integral
part of managing their network of critical devices.
ƒƒ Data centers are deploying rack PDUs capable of supplying multiple circuits, higher
voltages and higher currents.
ƒƒ Rack PDUs are available in a variety of rack form factors, that lead to neater cable
arrangements and better airflow along with more efficient device powering.
ƒƒ Basic PDUs are falling out of favor as intelligent PDUs can collect power and environment
information and warn of potential problems.
ƒƒ Circuit breakers used in rack PDUs are typically thermal-magnetic or hydraulicmagnetic with delay curves that allow for reasonable inrush currents while protecting
devices from excessive fault currents.
ƒƒ Typically, circuit breakers are available as single, double or triple pole devices. Singlepole breakers are optimal because they allow a single circuit to be de-energized.
ƒƒ Circuit breaker metering is important, particularly when dealing with high power
because tripping a breaker can be disastrous if it means losing several blade servers.
ƒƒ Rack PDU power cords may exit the PDU itself from the rear, the front, the top, or the
bottom. This will affect how the PDU fits into your rack and the bend radius of the cable.
ƒƒ Virtually all rack PDUs designed for data center use have built-in displays to show
current draw for the entire PDU unit.
ƒƒ To remotely access a rack PDU there are typically two choices: a graphical user interface
(GUI), or a Command Line Interface (CLI).
ƒƒ Integration with central directory services is important when the PDU offers the ability
to remotely turn on/off/recycle individual outlets or groups of outlets.
ƒƒ PDU manufacturers are offering environmental sensors to measure: rack air
temperature at the server inlets, humidity, airflow, vibration, smoke, water, and air
pressure.
ƒƒ The management system for data center power is often run on a “management network”
separate from the production network. This reduces the likelihood of a Denial of Service
(DOS) or other attack that would affect this critical function.
ƒƒ A rack PDU management system is a software application that consolidates all
communications with your rack PDUs. Its main functions are data collection, reporting,
power control, element management and fault management.
46
The iPDU Handbook
4.0 Considerations for Planning and
Selecting Rack PDUs
3.0 Elements of the System
47
4.0 Considerations for Planning and
Selecting Rack PDUs
4.1 Power Available and Distributed to Racks
4.2 Power Requirements of Equipment at Rack
4.3 Rack PDU Selection
4.4 Power Efficiency
48
The iPDU Handbook
T
here are several approaches to deploying power to IT equipment racks, which
affect rack PDU selection and configuration. Some approaches provide degrees
of redundancy and hence higher reliability/availability than others, but may not
be appropriate for certain types of equipment. Redundancy and higher availability
require resources, so managers of data centers that have limited power resources need
to decide what IT equipment justifies redundant power, e.g. production servers, and what
equipment does not, e.g. non-production equipment being tested or evaluated.
4.1 Power Available and Distributed to Racks
4.1.1 Single Feed to Single Rack PDU
The simplest power deployment to an IT
equipment rack is a single appropriately sized
power feed to a single rack PDU. IT equipment
with one or more power supplies would plug
into this single rack PDU. If that single feed
or single rack PDU should fail, for whatever
reason, power to the equipment in the rack
will be lost. The failure could occur at the
rack PDU itself or farther upstream (perhaps
a main feed fails or a building PDU circuit
breaker trips).
As noted earlier, the NEC requires that
circuits be loaded to no more than 80% of
their maximum capacity. For example, if a
30A feed and rack PDU were deployed in this
configuration, the load allowed (the rated
current) would be 24A (30A x 80%). NEC
would expect the feed and PDU to handle a
maximum of 30A but the circuit should be
loaded to only 24A.
4.1.2 Dual Feed to Single Rack PDU with
Transfer Switch
The next step up in availability is still a
Figure 17: Single Feed to Single Rack PDU
single feed to a single rack PDU, but with
(Source: Raritan, Inc.)
the addition of a transfer switch, which
typically has two feeds from the same or from
different building feeds. If one of the feeds to
the transfer switch fails, the transfer switch
automatically switches to the other power
feed and the rack PDU continues to supply power to the IT equipment. However, if the single
rack PDU fails, the power to the IT equipment is lost.
Learn more about transfer switches at www.raritan.com/TransferSwitch
4.0 Considerations for Planning and Selecting Rack PDUs
49
There are two types of transfer switches: static transfer switch (STS) and automatic transfer
switch (ATS). A STS is based on static electronic component technology (silicon controlled
rectifier or SCR), which results in faster and better controlled transfer between sources. An
ATS is less expensive and is based on electro-mechanical relay technology, which results in
slower transfer times.
Again, with this arrangement, the rack
PDU is still loaded to 80% of the maximum,
but the electrical power capacity required
has doubled – one feed is operational
and the second feed is a backup. It has
also doubled the amount of upstream
equipment necessary to supply the
additional feed.
Two power feeds to an ATS and then to a
single rack PDU is generally used only
where reliability is a concern and the IT
equipment itself, e.g. a server, has only one
power supply.
4.1.3 Dual Feed to Dual Rack PDUs
Today, many servers, network devices,
storage systems, even keyboard, video,
mouse (KVM) switches and serial console
servers, are available with dual power
supplies. Some larger servers may have as
many as four or even six power supplies.
The most reliable deployment here is to
use two power feeds to two rack PDUs.
With this configuration, if one rack PDU
or power feed fails there is a second one
available to maintain power to the IT
Figure 18: Dual Feed to Single Rack PDU
equipment in the rack. A common practice
with Transfer Switch
when using dual feeds is to use rack PDUs
(Source: Raritan, Inc.)
with colored chassis, such as red and
blue. The colored chassis enables a visual
control for installation of or changes to the PDU and connections. The rack will have a red
chassis PDU fed by input circuit “A” and a blue chassis PDU fed by input circuit “B”. The
colored chassis helps to eliminate confusion about which PDU is fed by circuit “A” or “B”.
It is important to remember the requirement that each circuit be loaded to no more than
40%. If the two circuits feeding the rack are both loaded to 80%, the NEC requirement will
be met, but think about what would happen if one of the circuits failed. The power demand to
the second circuit would jump from 80% to 160% and the circuit breaker for that feed would
trip so the second circuit to the rack would also lose power. To prevent this both feeds should
be loaded to no more than 40%, so that if one fails, the remaining circuit won’t be loaded
to more than 80%. Compared to the previous case with ATS (Section 4.1.2) where one feed
50
The iPDU Handbook
is the backup, in this configuration, both
feeds are powering IT equipment.
Note that if you intend to perform remote
switching for IT equipment with dual power
supplies, you will want to use a rack PDU
that supports outlet grouping; i.e., two or
more outlets are controlled as though they
are a single outlet.
4.1.4 Multiple Power Supplies
IT equipment with two or more power
supplies can vary in the way power is
delivered to the equipment. Some devices
have a primary and backup power supply;
some alternate between the power
supplies; and some devices share power
demand across all the power supplies. For
example, a blade server with four power
supplies in a 3+1 redundancy configuration
would draw one third of its power from
each of its three primary power supplies,
leaving one for redundancy in the event
any one of the three fail. Some more
sophisticated devices have multiple power
supplies, which are designed for both
redundancy and efficiency. For example,
Figure 19: Dual Feed to Dual Rack PDUs
some devices might drive utilization rates
(Source: Raritan, Inc.)
higher on specific power supplies to drive
higher efficiency. You will need to check
with each equipment manufacturer to understand how the power supplies work so that
optimal balanced load configurations can be achieved on the rack PDU, especially those
with branch circuits and 3 phase models.
4.1.5 Load Balancing
Load balancing is a procedure which attempts to evenly distribute the rack equipment’s
current draw among the PDU’s branch circuits so that as you come closer to perfect
balance, more total current can be supplied with the greatest headroom in each branch
circuit. For example, consider a PDU with two 20A circuit breaker protected branch circuits
– where each branch contains a number of outlets. The total current capacity of the PDU
is 40A with the limitation that no branch circuit of outlets can exceed 20A. If the total load of
all devices plugged into the PDU is 30A, perfect balance is achieved when the load is exactly
divided between the two branches (15A each branch). The headroom in each branch is then
5A (20A circuit breaker less 15A load). Any other distribution of the load (16A:14A, 17A:13A)
results in less headroom.
4.0 Considerations for Planning and Selecting Rack PDUs
51
Load balancing has similar benefits for three phase PDUs. As the load comes closer to
perfect balance, the current draw is more evenly distributed among the three phase lines
(more upstream headroom) and total current flowing in the three lines is minimized. For
example, consider a 24A three-phase delta wired PDU with three branch circuits. When an
18A load is balanced across the three branch circuits (6A load in each branch), the current
flowing in each input phase line is 10.4A (6 x sqrt 3 or 6 x 1.732) and the total current in all
three lines is 31.2A (10.4 x 3). If the entire load was carried by one branch circuit (totally
unbalanced), the current in the three phase lines are 18A, 18A, and 0A, respectively and
the total current is 36A. When the load is balanced across all three lines, the PDU has 7.6A
(18.0A – 10.4A) more headroom.
Load balancing can be tricky because many IT devices draw power in varying amounts
depending on the computational load. For devices with single power supplies an estimate of
the power consumption should be made for each device and then the devices plugged into
the circuits, so that the circuits are loaded evenly. This is true both within a rack and across
multiple racks. For devices with dual power supplies they should be plugged into different
circuits. A typical deployment would be the
dual feed to dual rack PDUs mentioned
above.
For IT devices with more than two
power supplies, such as blade servers,
load balancing can become even more
complicated, especially if the rack PDUs
are three-phase models. As an example,
assume four blade chassis are to be
installed in a rack, each chassis has six
power supplies and two three-phase rack
PDUs will be installed in the rack for
redundant power. The first blade server
will have power supplies (PS) # 1, # 2 and #
3 plugged into circuits ( C ) # 1, # 2 and # 3
respectively on PDU A and power supplies
(PS) # 4, # 5 and # 6 plugged into circuits
( C ) # 1, # 2 and # 3 respectively on PDU
B. Since we want to try to balance the load
across all circuits and lines, and we can’t
be sure that each of the four blade servers
will be performing tasks that equally load
the circuits, we will stagger the second
blade server power supplies. So the second
server will have PS # 1 plugged into C # 2;
PS # 2 plugged into C # 3; PS # 3 plugged
into C # 1 on PDU A; PS # 4 plugged in C
# 2; PS # 5 plugged into C # 3, and; PS # 6
plugged into C # 1 on PDU B. Circuit level
metering, phase level metering, and outlet
level metering will be very helpful for (re)
balancing loads in the rack.
52
Figure 20: Multiple Power Supply Configuration
(Source: Raritan, Inc.)
The iPDU Handbook
4.1.6 Inrush Current
Servers draw substantially more current when they are first turned on. This is known as
inrush current. As discussed in the section on overload protection, rack PDUs with circuit
breakers are designed not to trip during very short periods of high currents. However, it’s
easier on upstream circuits if the sudden surge when equipment is turned on is minimized.
For this reason, some rack PDUs provide outlet sequencing and allow users to configure
both the sequence and also the delay time in which the outlets are turned on. Some rack
PDUs even allow programming of outlet groups and allow sequencing of groups of outlets.
4.2 Power Requirements of Equipment at Rack
Section 4.1 deals with ways to deploy electrical power to a rack. This section deals with
determining how much power to deploy to a rack. Typically, the starting point is an IT device’s
nameplate power requirement data (see section 2.5.1) which specifies a voltage and current
(amps), which is typically higher than what is usually seen during actual deployment. As a
result, there is a convention of using a percentage of this value, e.g. 70%, when computing
the maximum PDU load capacity required: PDU load capacity = sum (device nameplate (V x
A)) x 70%). For example, 208V x 2.4A x 70% x 14 servers = 4.9kVA.
For the example above, if you run 208V you need a 30A (5kVA) rack PDU since you will load it
to 80% to meet North American requirements (4.9kVA / 208V = 23.5A, 23.5A is approximately
80% of 30A). If you want redundancy, add a second 5kVA rack PDU and load both PDUs up
to 40%. You will need to specify the appropriate number of outlets. It is a good idea to have
a few spare outlets for other devices, even if the rack PDU will be at its maximum capacity.
More efficient or different equipment might be installed in the rack in the future or servers
may not run near full capacity, leaving additional power capacity to power more equipment.
The current best practice is to standardize on IEC C-13 and/or C-19 PDU outlets, and 208V.
Most servers and data center devices can run at 208V (even up to 240V).
Remember the derating factor of 70% was just an estimate. Research has been done
with sophisticated rack PDUs that accurately measure power at the outlet. The findings
were surprising. Even at peak power consumption, 15% of the servers drew 20% or less of
their nameplate rating. Equally surprising was that nearly 9% drew 81% or more of their
nameplate rating. The point here is that the actual power consumed as a percentage of
nameplate rating can vary widely. Ideally, data center managers should measure the actual
power consumption rather than use a rule-of-thumb average such as 70%. If the actual
overall average is closer to 40%, as it was in the study, deploying power at 70% of nameplate
is wasteful and strands unused power.
If a cabinet populated with 30 1U servers has dual power feeds and the servers require
an average of 150W each, then the total power requirements for a cabinet are 150W x 30
servers = 4.5kVA. Assuming 250VA for additional equipment, like an Ethernet switch and
a KVM switch, this brings the total to 4.75kVA. So a 208V 30A PDU, which is rated at 5kVA,
would be sufficient. Such a PDU can carry the full load of 4.75kVA in a failover situation when
the power feed to one side of the cabinet fails, or is taken down for maintenance. Typically,
each PDU would be carrying only 40% of the 4.75kVA.
It is also important to note that three-phase Wye 208V rack PDUs are able to support
both 120V and 208V in the same PDU. This can be handy for situations where a variety of
equipment types with different voltage requirements need to be racked together.
4.0 Considerations for Planning and Selecting Rack PDUs
53
4.2.1 Rack PDU 208V Single-Phase vs. 208V Three-Phase
In a rack of 42 1U servers, if each server consumes an average of 200W then the total power
consumption is 42 x 200W = 8.4kW. To allow for the NEC requirement of 80% the rack needs
10.5kVA (8.4kW / 0.8). To allow for redundant power feeds two rack PDUs able to provide
10.5kVA are required. 208V single phase at 60A (48A rated) can deliver 10.0kVA. This could
suffice, particularly if the 200W per server estimate is on the high side. Another alternative
is 208V three phase at 40A (32A rated) which can deliver 11.5kVA. The 208V three-phase
alternative provides headroom to add higher power demand servers in the future and can
handle the existing servers even if their average power consumption increases from 200W
to 220W.
The use of three-phase power enables one whip or rack PDU to deliver three circuits instead
of just one circuit. The whip or input power cord on the rack PDU will be somewhat larger
for three-phase power than single phase power because instead of three wires (hot, neutral
and ground) a three-phase cable will have four or five wires.
The two three-phase alternatives are Delta and Wye. A three-phase Delta system will have
four wires – Line 1 (hot), Line 2 (hot), Line 3 (hot) and a safety ground. Individual circuits
are formed by combining lines. Three circuits are available – L1+L2, L2+L3 and L1+L3. The
power on each of the lines is a sine wave (this is also the case for single-phase power), but
each of the three sine waves is 120 degrees out of phase with the other two.
For three-phase power, the sine waves are 120 degrees out of phase, so calculating VA
is slightly more complex because we need to include the √3 which is 1.732. The apparent
power formula for three phase is V x Derated A x 1.732 = VA. As an example, three phase
208V, 40A (32A derated) is 208V x 32A x 1.732 = 11.5kVA. A three-phase Delta deployment
provides three separate circuits and over 70% more total power than a comparable single
phase, single circuit.
A three-phase Wye system will have five wires – Line 1 (hot), Line 2 (hot), Line 3 (hot), a
neutral and a ground. Individual circuits are formed by combining lines and by combining
a line with the neutral. As an example, a three-phase 208V Wye rack PDU supports three
208V circuits (L1+L2, L2+L3, L1+L3) and three 120V circuits (L1+N, L2+N, L3+N). Threephase Delta and three-phase Wye have the same apparent power, but the three-phase Wye
provides two different voltages while the three phase Delta only provides one voltage.
In North America, there may be a requirement for 120V convenience outlets, such as NEMA
5-15R (120V, 15A, 12A rated) or 5-20R (120V, 20A, 16A rated). These can be supported by
208V three-phase Wye PDUs where wiring between lines (L1, L2, L3) and lines and the
neutral can provide power to both 208V and 120V outlets. Whether the three-phase wiring
is Delta or Wye, the voltage is always referenced to the line-to-line voltage, not the line-toneutral voltage. This is even true in the 400V example on the next page (Section 4.2.2), where
all the outlets are wired line-to-neutral.
Since the Wye system adds a neutral wire, many data centers are wired for Wye and use
whips terminated with Wye receptacles, such as NEMA L21-30R. This means the data center
can use Wye PDUs that support 120V/208V or use Delta PDUs that support only 208V without
needing to change the data center wiring. A Delta PDU would use a NEMA L21-30P (the
mating Wye plug) but would not use a neutral inside the PDU. This is a perfectly acceptable
54
The iPDU Handbook
practice. For example, a data center could deploy Delta PDUs to racks where there is only a
need for 208V and Wye PDUs to racks where there is a need for both 120V and 208V.
Three-phase cables may be slightly larger than single-phase cables but it is important to
remember that one slightly thicker three-phase cable will be significantly smaller and weigh
less than three single-phase cables for the same voltage and amperage.
4.2.2 Rack PDU 400V Three-Phase
As shown in the 208V/120V example (Section 4.2.1), three-phase Wye wiring is a convenient
way to step down voltage. This is particularly true for 400V power. A generally accepted
method of delivering substantial power to densely packed racks is via 400V three-phase
Wye rack PDUs. 400V power distribution from panels to racks is now an accepted practice. A
data center designer could specify 400V Wye whips to 400V Wye rack PDUs. Since most data
center equipment can safely operate on voltages ranging from 100V to 240V, the 400V Wye
PDU can provide three circuits – L1+N, L2+N, L3+N – each supplying 230V (400V / 1.732).
400V Wye rack PDUs do not lend themselves to supporting 120V outlets as 208V Wye rack
PDUs do.
4.3 Rack PDU Selection
4.3.1 Rack PDU Selection and Special Application Requirements
There are many factors involved in selecting a rack PDU: data center location, application
requirements, IT equipment requirements, available power, space in the cabinet, energy
management, and efficiency objectives will combine to dictate what type of PDU should be
used. Some of the considerations below will guide you to select the feature set and hence
the type of PDU you will need to satisfy your requirements.
What type of equipment and how many devices are going into the cabinets, e.g. 42 x 1U
servers with a single feed per device, versus three 10U high blade servers with six power
supply feeds per server? The answer will help define the physical configuration, e.g. number
and type of outlets, and capacity of your PDU(s), e.g. how much power (kW) the PDUs need to
support. Average rack power requirements have risen from 6.0kW in 2006 to 7.4kW in 2009
and 12.0kW in 2011, and it is not unusual to see racks wired to provide as much as 30kVA.
Decision criteria for 24/7 manned sites will be different than remote management of lightsout facilities. If you need remote or lights-out management of a facility, then you will probably
need a switched PDU, which will require more security and user access management.
Remote applications may also call for SNMP management.
Integration with directory services, like LDAP or Microsoft’s Active Directory, is increasingly
a requirement for controlling access to resources, rather than requiring a separate
access control system. This capability is applicable to all applications, requiring central
authentication, local or remote. And for many data center applications, e.g. federal
government and financial institutions, encryption and strong password support are
necessary for remote access.
The rack PDU must supply uninterrupted power to each device plugged into it. You will want
4.0 Considerations for Planning and Selecting Rack PDUs
55
to prevent or mitigate any events that can potentially cause the circuit breaker on the rack
PDU or any point upstream to trip. Outlet sequencing is a valuable feature to prevent inrush
current from tripping a circuit breaker by establishing a sequence and appropriate delay for
powering multiple devices. Outlet sequencing not only prevents the undesired tripping of a
circuit breaker but also lets the user specify the order in which services (devices) come online
or are shut down during power cycling. For example, you will want to power the database
service before the web servers. This capability is most useful when used in conjunction with
the outlet grouping capability.
For some applications and equipment, you many need a customizable alarm threshold for
each outlet, with the capability to switch off an outlet should it exceed a certain power draw.
This would prevent a temperature or other sensor (see Section 3.2) from causing a shutdown
of servers. An advanced application is the control of the HVAC system using the temperature
reported by PDU temperature sensors.
In many mission critical environments, managed devices often have multiple feeds which will
be fed from different feeds or circuits for failover and redundancy. The managed device needs
to be managed as a single device regardless of the number of power supplies/plugs, and all
outlets must be handled simultaneously. This capability is applicable to all applications, local
or remote.
Event-driven power cycling of an outlet/device is required for some applications, particularly
for remote or unmanned sites. For example, if a device in a remote location fails to respond
and the WAN is not operational, there are basically two options: first, an expensive, timewasting truck roll to restart, or second, a rack PDU with the intelligence to trigger a restart
of a malfunctioning device, e.g. if the device has not responded for 20 minutes, recycle power
to the device.
If there is a need to maximize power efficiency, then rack PDUs can provide valuable data to
support those efforts: current, voltage, and power factor measurements at the PDU, line,
breaker, and outlet-level. Look for accurate kWh metering at the outlet level, especially if you
intend to report or charge back individuals or groups for usage. Metering accuracy can vary
significantly and, for some rack PDUs calculations may be based on assumptions and not
actual real-time measurements.
4.3.2 Rack PDU Functionality
Rack PDUs can vary significantly, not only in the operational functions they offer, but also in
their monitoring and data collection. Below is an overview of the strengths and weaknesses
of the four types/classes of rack PDUs previously defined in Section 2.1.1. Clearly, our class
definition is not rigid, since features offered by vendors will vary and you will want to select
PDUs based on the total fit to your requirement, but this can be a useful guide in your selection.
Non-Intelligent PDUs
ƒƒ Basic PDUs - are power strips that are used in critical environments such as data centers.
They distribute correct voltage and current to multiple outlets to power IT equipment in
racks.
ƒƒ Monitored PDUs - allow a user to view a local display that typically provides electric
current only. However, this information cannot be accessed remotely as the units have no
network connectivity capabilities.
56
The iPDU Handbook
Intelligent PDUs
ƒƒ Metered Input PDUs - Meter power at the PDU-level, and can display the data both
locally and over a network. Metering helps users to determine power usage and available
capacity at the rack, and facilitates provisioning. By metering at the input level, users
can avoid overloading circuits and more easily calculate efficiency metrics like Power
Usage Effectiveness (PUE).
ƒƒ Metered Outlet PDUs - Meter power at the outlet-level, and can display the data both
locally and over a network. Like metered input PDUs, outlet-metered models help
users to determine power usage and available capacity at the rack, and facilitates
provisioning. Most importantly, outlet-level metering allows users to understand power
consumption at the device or server-level which make it possible to allocate costs to
specific business units or customers.
ƒƒ Switched PDUs - Offer the features of Metered Input PDUs and also provide controlled
on/off switching of individual outlets. They enable authorized users to securely power
cycle devices remotely in a specific order, offer power sequencing delay to minimize
inrush currents, prevent unauthorized device provisioning, and can power off devices
that are not in use to conserve energy.
ƒƒ Switched PDUs with Outlet Metering - Combine all of the capabilities of Switched PDUs
with those of Outlet-Metered PDUs.
4.3.3 Benefits of an Intelligent Rack PDU
Over the past decade, the IT industry has chosen to move to more sophisticated, manageable
systems. This fact is nowhere more evident than the trend toward the use of intelligent
PDUs.
An intelligent PDU will provide, at minimum, real-time outlet-level and/or PDU-level power
monitoring, rack temperature and humidity monitoring, and may provide remote outlet
switching. For top-tier data centers, deployment of intelligent PDUs can make a significant
difference in the ability of IT administrators and facilities managers to improve uptime
and staff productivity, efficiently utilize power resources, make informed capacity planning
decisions, and save money. And, in so doing, they will operate greener data centers.
Clearly, if your data center has dozens of racks, then the greatest benefits will be realized
by using a rack PDU management system to consolidate data acquisition, reporting, PDU
administration, and control. Let’s examine the benefits of deploying intelligent PDUs in
depth.
Improve Uptime and Staff Productivity
ƒƒ Monitoring power at a PDU and individual outlet level, with user-defined thresholds and
alerts via e-mail or SNMP, provides awareness of potential issues before they occur.
ƒƒ Remote reboot of servers and IT equipment from anywhere in the world via a web
browser reduces downtime and personnel costs.
Use Power Resources Safely
ƒƒ User configurable outlet-level delays for power sequencing prevent circuits from
tripping from IT equipment in-rush currents.
4.0 Considerations for Planning and Selecting Rack PDUs
57
ƒƒ Control of outlet provisioning prevents accidentally plugging IT equipment into circuits
that are already heavily loaded and are at risk of tripping circuit breakers.
Make Informed Power Capacity Planning Decisions
ƒƒ Outlet-level monitoring may identify some simple rearrangements of equipment to free
up power resources by balancing power demands across racks.
ƒƒ Monitoring power at the outlet-level can identify equipment that may need to be changed
to stay within the margin of safety of defined thresholds.
ƒƒ Monitoring rack temperature and other environmental conditions can prevent problems,
especially when a data center is rearranged and airflow patterns change.
Save Power and Money
ƒƒ Monitoring power at the outlet-level combined with trend analysis can identify ghost or
underutilized servers that are candidates for virtualization or decommissioning.
ƒƒ Remote power cycling enables IT managers to quickly reboot hung or crashed IT
equipment without incurring the cost of site visits.
ƒƒ Temperature and humidity sensors help data center managers optimize their air
conditioning and humidity settings and avoid the common practice of overcooling and
related waste of energy.
4.4 Power Efficiency
4.4.1 Introduction to Power Usage Effectiveness (PUE)
It is important to understand the role the rack PDU can play in providing accurate data
that IT and facilities management will use to improve efficiency. We all know the maxim,
“You can’t manage what you don’t measure,” but be aware that unless the data is clearly
understood, it can unwittingly lead to false conclusions and inappropriate actions.
The objective is to maximize the amount of useful computational work by IT equipment for
total energy consumed by the equipment itself and the infrastructure that supports the
equipment. The most commonly used metric is The Green Grid’s Power Usage Effectiveness
or PUE:
PUE = Total Facility Power ÷ IT Equipment Power
“Total Facility Power” in this equation is all the power required to operate the entire data
center, including the IT equipment items: servers, storage, network equipment and other
IT equipment; and the support infrastructure items: CRAC units, fans, condensers, UPS
and lighting. “IT Equipment Power” is the power required to operate the servers and IT
equipment alone. Theoretically PUE can range from 1.0 (where all the power is consumed
by IT equipment only) to infinity. A PUE = 2.0 means that IT equipment is consuming 50% of
data center power. Another commonly used metric is Data Center Infrastructure Efficiency
(DCiE), which is the inverse of PUE. Since they are derived from the same data, there is no
substantive difference in the measurement or its usage.
58
The iPDU Handbook
A recent Lawrence Berkeley National Laboratory study of data center power allocation
showed an average data center PUE of 2.18 and a similar result was found by EYP Mission
Critical Facilities, Inc. (now an HP company), with the following allocation:
Figure 21: Data Center Power Consumption
(Source: EYP Mission Critical Facilities, Inc.)
4.4.2 PUE Levels
The Green Grid defines three levels of PUE: Basic or Level 1, Intermediate or Level 2, and
Advanced or Level 3. Many industry analysts recommend measuring IT power consumption
at the Intermediate, Level 2; i.e., at the PDU level. While it is true that PDU-level power
consumption will provide the denominator needed to calculate PUE, this information alone
is likely insufficient to drive the best efficiency improvement decisions. Regardless of the
PUE Level you choose to employ, the best practice is to gather data over a time period of
“typical’ power usage to ensure that the peaks and valleys have been captured in order to
establish a baseline and to track your improvements.
Level 1
Basic
Level 2
Intermediate
Level 3
Advanced
IT equipment power
measurement from
UPS
PDU
Server
Total facility power measurement
from
Data center input power
Data center input less
shared HVAC
Data center input less
shared HVAC plus
building, lighting,
security
Minimum measurement interval
Monthly/weekly
Daily
Continuous
Figure 22: Definition of Green Grid PUE Levels
(Source: The Green Grid)
4.0 Considerations for Planning and Selecting Rack PDUs
59
4.4.3 Why Advanced, Level 3 PUE?
An improved (lower) PUE may be misleading since it can result from inefficiencies in the
power consumed by IT equipment which merely increases the denominator. A lower PUE
is generally better than a higher one, but it is possible to implement measures that reduce
data center energy consumption yet actually increase your PUE. For example, if you were to
replace older, less efficient servers with more efficient ones, eliminate ghost servers, turn
off servers that were idle during the night, or employ server virtualization, the net result
would be power reduction, but your PUE would actually increase. The detailed IT load data
from Level 3 provides the granularity of information to reduce energy consumption, not just
improve the PUE metric. Clearly, PUE (and its inverse, DCiE) becomes a more useful beacon
once you have built efficiency into the IT equipment performance; to do that you will want
the granular power usage data for the Advanced, Level 3 PUE metric. Then you can attack
the numerator and squeeze inefficiencies out of the infrastructure.
Backup
Generator
2 MW
A
AC
Utility Feed
Substation
15 kV - 480 V/277 V
L1
ATS
2500A
ATS
2500A
Automatic
Transfer
Switch
Automatic
Transfer
Switches
ATS
2500A
UPS
B Side
320 kW
400 kVA
UPS
B Side
320 kW
400 kVA
UPS
B Side
320 kW
400 kVA
UPS
B Side
320 kW
400 kVA
L1
L1
L1
L1
ATS
2500A
L2
L2
Mechanical
Switch Gear
Mechanical
Switch Gear
Mechanical
Equipment
B
Mechanical
Equipment
L 3 L 3 L 3 L 3
L 3 L 3 L 3 L 3
Building
Building
Management
Management
Facility Power
Facility Power
CRAC Units
CRAC Units
Glycol Pumps
Glycol Pumps
Maintenance Bypass Switch, 2500 A, 480 V
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
PDU
100
kVA
L2
L2
L2
L2
L2
L2
L2
L2
L2
L2
L2
L2
L2
L2
Telco
L 3
Redundant Feeds
For IT Equipment
Backup
Generator
2 MW
Generator Switchgear
Standby Switchboard, 2500 A, 480 V
Panel, 800 A, 480 V
Backup
Generator
2 MW
Backup
Devices
Storage
L 3
L 3 Security
Encryption
KVM
L 3 Console
L 3
Servers
Switches L 3
Routers L 3
Printers L 3
PC’s & L 3
Workstations
Key:
L1 = Level 1
Uninterruptible
Power Supply (UPS)
L2 = Level 2
Power Distribution
Unit (PDU)
L3 = Level 3
Individual Device
L 3
Figure 23: Measuring at the Three PUE Levels
(Source: The Green Grid)
60
The iPDU Handbook
4.4.4 The Advantages of High Power
A single-phase 120V at 100A (80A rated) circuit provides 9.6kVA. A single-phase 208V at 60A
(48A rated) circuit provides 10.0kVA. A three-phase 208V at 40A (32A rated) circuit provides
11.5kVA. A single-phase 230V at 60A (48A rated) circuit provides 11.0kVA. A three-phase
400V at 20A (16A rated) circuit provides 11.1kVA.
Running higher voltages at lower currents means smaller cables which use less copper,
weigh less, take up less space and cost less. Running three-phase power instead of singlephase power means fewer cables which simplifies deployment, and provides the benefits of
smaller cables, less copper, less weight, and less cost.
Plugs and receptacles are also less expensive at higher voltages and lower current ratings.
For example, a 30A 400V three phase Wye (16.6kVA) plug (Hubbell NEMA L22-30P) costs
$32 and the receptacle costs $41. A 60A 208V three phase Delta (17.3kVA) plug (Mennekes
IEC309 460P9W) costs $166 and the receptacle costs $216. The plug/receptacle combination
is $73 vs. $382.
There are other benefits to higher voltages. A 400V power circuit will eliminate voltage
transformations, and can reduce energy costs by approximately 2-3% relative to 208V
distribution, and approximately 4-5% relative to 120V distribution.
Consolidating data centers will generally reduce total power consumption and may create
opportunities for the use of high-density racks and high power rack PDUs. For example, a
42U rack filled with 1U servers consuming 250W each draws 10.5kW which would require
two three phase 208V, 50A circuits providing 14.4kVA each. Taking advantage of blade
servers might lead to deploying four blade chassis in one rack which would require two
three phase 208V, 80A or two three phase Wye 400V, 50A rack PDUs. These examples allow
sufficient headroom should one of the feeds fail. They also support the North American
requirement for 80% derating.
High-density racks can be effectively deployed in small, medium, or large data centers. Even
small data centers will benefit from racks delivering high power for multiple blade servers
or densely packed 1U servers.
4.0 Considerations for Planning and Selecting Rack PDUs
61
Chapter 4 Summary
ƒƒ The simplest power deployment to an IT equipment rack is a single appropriately-sized
power feed to a single rack PDU.
ƒƒ The next step up in availability is still a single feed to a single rack PDU, but with the
addition of a transfer switch, that typically has two feeds from the same or different
building feeds.
ƒƒ The most reliable deployment is to use two power feeds to two rack PDUs, so if one
rack PDU or power feed fails there is a second one available to maintain power to the IT
equipment in the rack.
ƒƒ IT equipment with two or more power supplies can vary in the way power is delivered to
the equipment, often requiring load balancing.
ƒƒ Load balancing attempts to evenly distribute the rack equipment’s current draw among
the PDU’s branch circuits so that as you come closer to perfect balance more total
current can be supplied with the greatest headroom in each branch circuit.
ƒƒ Determining how much power should be deployed to a rack requires you to know device
voltage and ideally, actual current draw.
ƒƒ The use of three-phase power enables one whip or rack PDU to deliver three circuits
instead of just one circuit. One slightly thicker three-phase cable will be significantly
smaller and weigh less than three single-phase cables for the same voltage and
amperage.
ƒƒ There are many factors involved in selecting a rack PDU: data center location, application
requirements, IT equipment requirements, available power, space in the cabinet, energy
management, and efficiency objectives.
ƒƒ Deployment of intelligent PDUs can make a significant difference in the ability of IT
administrators and facilities managers to improve uptime and staff productivity,
efficiently utilize power resources, make informed capacity planning decisions, and save
money.
ƒƒ The most commonly used metric for data center power efficiency is The Green Grid’s
Power Usage Effectiveness or PUE. The best practice is to gather data over time of
“typical’ power usage to ensure that the peaks and valleys have been captured in
calculating your PUE to establish a baseline and to track your improvements.
ƒƒ Running higher voltages at lower currents means smaller cables which use less copper,
weigh less, take up less space and cost less. Running three-phase power instead of
single-phase power means fewer cables which simplifies deployment as well as the
benefits of smaller cables, less copper, less weight and less cost.
ƒƒ High-density racks can be effectively deployed in small, medium, or large data centers.
Even small data centers will benefit from racks delivering high power for multiple blade
servers or densely packed 1U servers.
62
The iPDU Handbook
5.0 Future Trends for Rack PDUs
4.0 Considerations for Planning and Selecting Rack PDUs
63
5.0 Future Trends for Rack PDUs
5.1 Higher Density, Higher Power Rack PDUs with Sensors
5.2 Increased Intelligence at the Rack to Support Efficiency Initiatives
5.3 Integration with Higher Level Data Center Management Systems
64
The iPDU Handbook
T
wo primary forces are influencing rack PDU development and innovation trends.
First is the demand for increasing power and density of IT equipment at the rack
or compute density per U of rack space. Second is the industry-wide goal, even
mission, to create energy efficient (often called “green”) data centers, including
carbon footprint reduction. Both trends challenge PDU vendors to improve both hardware
and software design; and the second requires all IT and facilities organizations to better
understand how the data center power is consumed and take active measures to reduce it.
The above trends are underscored in the IHS November 2013 report “The World Market for
Rack Power Distribution Units – 2013 Edition.” Compound Annual Growth Rate (CAGR), for
rack PDUs between 2012-2017 is predicted to be:
ƒƒ <5kW are expected to have unit growth of 4.4% (2.0% for the Americas).
ƒƒ 5kW – 10kW are expected to have unit growth of 5.9% (5.4% for the Americas).
ƒƒ >10kW are expected to have unit growth of 6.4% (7.5% for the Americas).
When comparing the growth rate of single-phase PDUs to three-phase PDUs, analysts
predict:
ƒƒ Single-phase PDU units are expected to grow 4.4%.
ƒƒ Three-phase PDU units are expected to grow 7.9%.
When comparing the growth rate of non-intelligent PDUs to intelligent PDUs, analysts
predict:
ƒƒ Non-intelligent rack PDUs are expected to have unit growth of 4.1%.
ƒƒ Intelligent rack PDUs are expected to have unit growth of 8.4%.
The factors that are driving PDU demand are increasing power consumption and higher
densities along with increasing the need for PDU intelligence.
What is the Power Density (in kW) per Rack in Your Data Center?
20
16.5
16
12
12
8
6
7.3
6.1
7.4
4
0
Spring 2006
Spring 2007
Spring 2008
Spring 2009
2011 Forecast
2019 Forecast
Data Center Users’ Group (DCBG) - Emerson Network Power
Figure 24: Average Power per Rack
(Source: Emerson Network Power)
5.0 Future Trends for Rack PDUs
65
Heat Load Per Product Footprint (watts / equipment sq.ft.)
10,000
8,000
6,000
4,000
2,000
1,000
800
600
400
200
100
60
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
Year Of Product Announcement
© 2005 ASHRAE TC 9.9 Datacom Equipment Power Trends & Cooling Applications
Figure 25: Equipment Power Density Trends
(Source: ASHRAE)
5.1 Higher-Density, Higher-Power Rack PDUs with Sensors
The growing popularity of 1U servers, blade servers, network-attached storage, storage
area networks (SANs) and multi-gigabit, chassis-based network communications gear,
place enormous demands on rack PDUs. For example, 4 blade server chassis in a single
rack could draw in excess of 20kW of power creating both power and cooling challenges
for data center managers. From a power perspective, racks will require three-phase power
with 60A, 80A, even 100A of service. There are some data centers bringing 400 volt threephase service to the rack, to accommodate this power demand while increasing efficiency
from reduced voltage step downs. Similarly, end users are packing dozens of 1U servers
into a single rack and pressing rack PDU vendors to design PDUs to support 40+ outlets
and 20+ kW.
Server virtualization is another major trend in data centers and should lead to improved
efficiency and cost reduction. However, running multiple virtual machines on one server
will drive up its total power consumption; and a rack containing several such servers
could potentially experience a lot more power consumption, thereby driving the need to
gain additional visibility into power loads of servers in the rack to optimally manage power
capacity.
More power consumption means more cooling to remove the additional heat. PDU vendors
will be expected to supply the basic environmental sensors for heat, humidity, and air flow
to help understand the overall environmental conditions and to identify zones that must be
fine-tuned or supplemented with dedicated or specialized cooling.
66
The iPDU Handbook
5.1.1 Customizing IT Equipment for Power Efficiency
One trend to watch, especially for the largest data centers, is the design and deployment of
custom servers, power supplies, rack PDUs, etc. to maximize power usage efficiency. For
example, Facebook along with OpenCompute have begun to deploy 480V three-phase Wye
power where each line is wired to the neutral so the outlets deliver 277V. This approach is
very efficient but highly specialized since most IT equipment today is not built with power
supplies that support 277V. Furthermore, common data center receptacles are IEC C-13 and
C-19, which do not support 277V.
The savings and efficiencies (1% - 2% over 400V/230V three-phase systems) are enough
that Facebook/ OpenCompute can justify building custom triplet racks, custom servers with
custom power supplies, custom battery/UPS, and 480V/277V rack PDUs with custom Tyco
3-pin Mate-N-Lock outlets.
5.2 Increased Intelligence at the Rack to Support Efficiency Initiatives
Many data centers have grown larger and more complex in recent years as the consolidation
trend continues. With increasing size and complexity there is a greater need to drive
intelligence to the IT equipment at the rack to create what industry people are beginning to
think of as the “Intelligent Rack.”
Every data center, regardless of size, is designed to support the servers at the rack where the
actual computing is taking place. It is also where the vast majority of the power is or should
be consumed. Proper monitoring and metering of the IT equipment along with environmental
sensors at the rack will collect the data necessary to produce the most significant overall
efficiency, savings and operational improvement. Collection and analysis of actual energy
data will enable you to maximize the use of current resource capacity and take advantage
of capacity planning tools to “right-size” the data center for future requirements. This will
allow you to eliminate or defer capital expenses of data center expansions; while improving
day-to-day energy efficiency and overall IT productivity.
Capacity planning based on nameplate data is no longer sufficient. Efficiency improvement
is an information-driven activity. In order to formulate and drive the most effective decisions,
you will need to collect IT device CPU utilization and their corresponding actual power usage.
More energy efficiency will be gained if such planning is based upon the trends observed
from the actual data over time. Furthermore, the actual data collected at the rack level
can be integrated with the overall data center infrastructure management (DCIM) systems
and data center energy management systems for complete data center and power chain
visualization, modeling, and planning, which can lead to further improvements in the data
center ecosystem, e.g. computing carbon emissions generated by IT devices to report on
and take steps to lower your carbon footprint.
Efficiency can also be gained from software that offers policy-based power control to
automatically turn servers on and off based on granular power consumption data and a set
of pre-established static or even dynamic rules. These power saving applications are finding
early adopters in development labs, web server farms and cloud computing environments.
Eventually they may find their way into mainstream data centers and their success will be
based to some extent on the deployment of intelligent PDUs to enable their functions.
5.0 Future Trends for Rack PDUs
67
Creating energy-efficient behavior throughout your organization is a key factor in reducing
waste and costs; and the essential ingredient to affect behavior is individual awareness and
accountability for energy usage. Of course, to be effective, any such energy reporting or
charge-back system must be based on credible, comprehensive, and coherent usage data,
so PDU vendors will be expected to deliver the highest accuracy for energy usage at every
level of the organization.
5.3 Integration with Higher Level Data Center Management Systems
In recent years, a variety of software products have been introduced to help both IT and
Facilities people better manage the data center. While the category name may differ—
Physical Infrastructure Resource Management (PIRM), Data Center Infrastructure
Management (DCIM), Data Center Service Management (DCSM)—these applications provide
most of the following major functions: database of all physical data center assets with
detailed data for IT, power and HVAC equipment, physical data center layout, and cable
connections; change management; 2D or 3D visualization of the data center building with
drill down to lowest level data element; and capacity planning based on availability of floor
and rack space, power, cooling, etc.
The data required to manage data center infrastructure and energy effectively are collected
from power devices along the entire power chain, from IT devices themselves, environmental
sensors, data center layout maps, cable plans, and cooling system design documents. The
more data collected, the more complete the information will be, and the better equipped
data center personnel will be to support critical IT operations reliably, efficiently, and costeffectively.
68
The iPDU Handbook
The following is a simplified view of data measurement, collection, compilation,
analysis and correlation and decision support:
ƒƒ Intelligent rack PDUs measure essential power data at a predefined frequency and store
such data in memory.
ƒƒ The data collection services from the rack PDU management (or power management)
system polls the intelligent PDUs through industry standard management protocols
such as SNMP.
ƒƒ The data collection service can be part of the intelligent PDU vendor’s rack PDU
management system, or it can be part of a DCIM or energy management system. For
scalability reasons, data collection is typically delegated to specific PDU vendor’s PDU
management system, which is deployed along with intelligent PDUs to administer,
maintain and troubleshoot the PDUs; as well as to collect power statistics from these
intelligent PDUs.
ƒƒ The rack PDU management system can use the collected data to perform the first level
of analysis. This will help to reveal power trends and pinpoint any potential issues. The
amassed data can be used by the DCIM or another energy management system for
further analysis.
ƒƒ The energy management or DCIM system provides visibility beyond the scope of the PDU
management system including: poll information from upstream smart power devices,
physical layout, cable plan, HVAC deployment information, etc. making it more suitable
for analysis that must take into consideration many more factors beyond the intelligent
PDUs.
With the advanced analysis conducted by a DCIM or energy management system, data
center management staff can make their day-to-day operational decisions as well as longer
term strategic planning, to provide reliable power for business applications while reducing
waste in data center energy consumption.
5.0 Future Trends for Rack PDUs
69
Chapter 5 Summary
ƒƒ Two factors influencing rack PDU development and demand are the increasing power
and density of IT equipment at the rack, and the industry-wide goal of creating energy
efficient data centers. These are driving the need for PDU intelligence.
ƒƒ The growing popularity of 1U servers, blade servers, network-attached storage, storage
area networks (SANs) and multi-gigabit, and chassis-based network communications
gear, place enormous demands on rack PDUs and create both power and cooling
challenges.
ƒƒ More power consumption means more cooling to remove the additional heat. PDU
vendors will be expected to supply environmental sensors for heat, humidity and air
flow to help understand the overall environmental conditions and to identify zones that
must be fine-tuned or supplemented with dedicated or specialized cooling.
ƒƒ Many data centers have grown larger and more complex in recent years as the
consolidation trend continues. With increasing size and complexity there is a greater
need to drive intelligence to the IT equipment rack to create what industry people are
beginning to think of as an “Intelligent Rack.”
ƒƒ Proper monitoring and metering of the IT equipment along with environmental sensors
at the rack will collect the data necessary to produce the most significant overall
efficiency, savings and operational improvement.
ƒƒ An essential ingredient of energy efficient behavior is individual awareness/accountability
for energy usage. Billing grade accuracy and billback reports can be driven by intelligent
PDUs and energy management software.
ƒƒ Data collected at the rack-level can be integrated with a data center infrastructure
management (DCIM) system for complete data center and power chain visualization,
modeling, and planning, which can boost efficiency and lower carbon footprint.
70
The iPDU Handbook
Reference
Reference
71
Reference
Key Terms to Know
References and Further Reading
Index
72
The iPDU Handbook
Key Terms to Know
4-wire: A 4-wire rack PDU consists of one ground wire and three lines (see Three phase
Delta below), each line carrying equal voltage but each voltage sine wave is 120 degrees
out of phase with the others. The voltage of two lines is available (line to line, e.g., L1-L2).
5-wire: A 5-wire system is the same as the 4-wire system but with the addition of a neutral
wire (see Three phase Wye below) so that the voltage of one line can be supplied (line to
neutral) as well as the voltage of two lines (line to line).
Active Directory®: is a directory service that authenticates and authorizes all users and
computers in a Windows domain type network—assigning and enforcing security policies
for all computers, and installing or updating software.
Active Power: The real power drawn by an IT device. This determines the actual power
purchased from the utility company and the heat loading generated by the equipment since
for IT equipment, 1W of electricity equals 1W of heat.
Air Pressure: Air pressure that is too high will result in both higher fan costs and greater
leakage which can short circuit cooling air, while pressure that is too low can result in hot
spots at the areas most distant from the cool air supply point.
Airflow: There are two primary areas for monitoring airflow in the data center: above the
floor and below the floor. Differential airflow sensors are used to ensure that the pressure
differential between the subfloor and the floor is sufficient to control air flowing from
the subfloor to the floor above. Blockages in under-floor supply plenums can cause high
pressure drops and uneven flow, resulting in cold spots in areas where cooling air is short
circuiting to the return path.
Amperage: Also known as the current. It is the flow or rate of flow of electrons, ions, or holes
in a conductor or medium between two points having a difference in potential, measured in
amperes and equal to the ratio of the voltage to the resistance.
Apparent Power: Expressed in volt-amperes (VA). The product of the voltage applied to the
IT device times the current drawn by the equipment. The VA rating is used for sizing wiring
and circuit breakers. The apparent power is always equal to or larger than the active power.
ASHRAE: The American Society of Heating, Refrigerating, and Air-Conditioning Engineers
(ASHRAE). ASHRAE recommends temperature probes be placed at specific locations at the
inlets of equipment rack to monitor temperatures.
Basic PDU: Basic PDUs are power strips that are constructed out of high-quality components
for use in critical environments such as data centers. They distribute correct voltage and
current to multiple outlets.
Branch Circuit: Branch circuits are power feeds that originate from a panel, switch or
distribution board and terminate into an electrical receptacle mounted in a junction box
near the IT equipment rack.
Reference
73
Circuit Breaker: An automatically operated electrical switch designed to protect an
electrical circuit from damage caused by overload or short circuit. Circuit breakers are
available on several PDU models.
Delay Curve: Gives a circuit breaker the ability to discriminate between normal and
damaging over-currents, which are quite common when powering servers.
Delta: This configuration gets the name Delta because a schematic drawing of it has three
transformers forming a triangle or the Greek letter Delta. The three lines connect to the
three “corners” of the triangle.
Derated: In North American data centers must be certified to UL 60950-1. UL 60950-1 limits
a device to draw no more than 80% of the rating of its input plug. This 80% limitation is
commonly known as “derated” current.
Energy Management: Driven by industry-wide goal of creating energy efficient data centers
and reducing carbon footprint. Requires information that leads to smarter energy usage and
more energy efficient behavior.
Fuses: A type of low resistance resistor that acts as a sacrificial device to provide overcurrent
protection of either the load or source circuit. Not ideal for data center energy distribution as
it may increase the mean time to repair (MTTR).
Ground: A conducting body, such as the earth or an object connected with the earth, whose
potential is taken as zero and to which an electric circuit can be connected. The purpose of
a ground wire is to safely direct stray currents to ground rather than allowing them to pass
through someone contacting the stray currents.
Humidity: A measurement of moisture in the air. High humidity can cause condensation
buildup on computer components, increasing risks of shorts. Likewise, if the humidity is too
low, data centers can experience electrostatic discharge (ESD).
IEC: The International Electrotechnical Commission creates classifications for plugs and
receptacles used in data centers.
Inline Meter: Device that allows users to monitor power remotely at the rack PDU inlet or
circuit-level.
Intelligent PDU: PDUs that can be accessed and controlled remotely via a web browser
or command line interface (CLI). They meter power at both the PDU and individual outletlevel; support alerts based on user-defined thresholds; provide security in the form of
strong passwords, authentication, authorization and encryption; and incorporate rich
environmental management capabilities.
Kilowatt-Hour (Kwh): A kilowatt-hour (kWh) is a unit of electrical energy or work, equal to
the power supplied by one kilowatt for one hour.
Lightweight Directory Access Protocol (LDAP): An open, vendor-neutral, industry standard
application protocol for accessing and maintaining distributed directory information services
over an Internet Protocol (IP) network.
74
The iPDU Handbook
Line: An electrical conductor which is a source of voltage, e.g., 120V. In a single-phase
system there are one or two lines. In a three-phase system there are three lines.
Load Balancing: A procedure which attempts to evenly distribute the rack equipment’s
current draw among the PDU’s branch circuits so that as you come closer to perfect balance,
more total current can be supplied with the greatest headroom in each branch circuit.
Metered PDU: PDUs that measure the current draw (load) at the PDU level, and display the
data locally. More sophisticated models also offer user-defined alarm functions and remote
access to the data over a serial or network port.
Nameplate: Nameplate data is the electrical power consumption information specified by
the equipment manufacturer. It is typically a conservative estimate of the maximum amount
of power the device could draw.
NEC: The National Electrical Code is a regionally adoptable standard for the safe installation
of electrical wiring and equipment in the United States. It has specific requirements for data
centers.
NEMA: The US National Electrical Manufacturers Association is the organization that sets
standards for electric outlet and receptacles found in data centers.
Neutral: An electrical conductor which provides a return path for the voltage supplied by a
line. The neutral itself is not a source of voltage, e.g., 0V.
Power Distribution Unit (PDU): Large data center PDUs are used earlier in the power chain
and take the form of panel boards mounted on walls or free standing pedestals. In this
handbook we’re discussing only the rack PDU, at the end of the chain, which supplies power
to the IT equipment in the rack.
Power Factor: Defined as the ratio of the real power flowing to the load, to the apparent
power in the circuit, and is a dimensionless number between -1 and 1.
Power Usage Effectiveness (PUE): A measure of how efficiently a data center uses energy;
specifically, how much energy is used by the computing equipment. Defined as PUE = Total
Facility Power / IT Equipment Power.
Rack PDU: A rack PDU is mounted in an IT equipment rack and provides electrical power to
various IT devices such as servers, networking, and storage equipment (See Section 2.1.1).
Rated: The Rating of an electrical appliance indicates the voltage at which the appliance is
designed to work and the current consumption at that voltage.
Simple Network Management Protocol (SNMP): An internet-standard protocol for
managing devices on IP networks. It is used mostly in network management systems to
monitor network-attached devices for conditions that warrant administrative attention.
Reference
75
Single-Phase: The distribution of alternating current electric power using a system in
which all the voltages of the supply vary in unison.
Switched PDU: Switched PDUs offer the features of metered PDUs and also provide
controlled on/off switching of individual outlets and load metering (see metered PDUs) at
the PDU level. They enable authorized users to securely power cycle devices remotely; and
they may also provide a power sequencing delay as well as some outlet use management.
Temperature: In the data center, temperatures must be kept at predetermined set points
in order to keep sensitive IT equipment from overheating, and prevent costly overcooling.
Three-Phase: Three-phase electric power systems have at least three conductors carrying
alternating current voltages that are offset in time by one-third of the period.
Underwriters Laboratories (UL): A safety consulting and certification company that drafts
several standards that affect electricity in the data center.
Voltage: Electromotive force, or difference in electrical potential, measured in volts and
equal to the current times the resistance.
Volt-Amp (VA): The product of the voltage applied to the IT device times the current drawn
by the equipment. The VA rating is used for sizing wiring and circuit breakers. The apparent
power is always equal to or larger than the active power.
Watt: The real power drawn by an IT device. This determines the actual power purchased
from the utility company and the heat loading generated by the equipment since for IT
equipment, 1W of electricity equals 1W of heat.
Watt-Hour (Wh): A unit of energy equivalent to one watt of power expended for one hour of
time.
Wye: An electrical configuration that derives its name from its schematic drawing of three
transformers meeting in the center forming the letter “Y”. The three lines connect to the
three “branches” of the “Y” and the neutral connects to the center of the “Y.”
76
The iPDU Handbook
References and Further Reading
Alger, Douglas. “Build the Best Data Center Facility for Your Business: A comprehensive guide to designing and
operating reliable server environments,” Indianapolis (IN): Cisco Press, 2005.
ASHRAE., “Thermal Guidelines for Data Processing Environments,” 2nd ed. Atlanta (GA): ASHRAE, 2009.
ASHRAE,. “ASHRAE Workshops on Improving Data Center Energy Efficiency and Best Practices,” NYSERDA
sponsored workshop; New York (NY) ASHRAE, November 6, 2009.
ASHRAE., “2008 ASHRAE Environmental Guidelines for Datacom Equipment,”. Atlanta (GA): ASHRAE, 2008.
ASHRAE., “Best Practices for Datacom Facility Energy Efficiency,” 2nd ed. Atlanta (GA): ASHRAE, 2009.
ASHRAE,. “High Density Data Centers: Case studies and best practices,”. Atlanta (GA): ASHRAE, 2008.
Haas, Jon, Monroe M, Pflueger J, Pouchet J, Snelling P, Rawson A, Rawson F. et al,. “Proxy Proposals for
Measuring Data Center Productivity,”. White Paper #14,. Beaverton, (OR): The Green Grid, 2009.
Raritan Inc., “Data Center Power Overload Protection: Circuit breakers and branch circuit protection for data
centers,”. White Paper,. Somerset (NJ): Raritan Inc., Somerset, NJ, 2009a.
Available at http://www.raritan.com/high-power-white-paper. Accessed February, 2014.
Raritan Inc., “Data Center Power Distribution and Capacity Planning: Understanding what you know – —and
don’t know – —about power usage in your data center.,” White Paper,. Somerset (NJ): Raritan Inc., Somerset, NJ,; 2009b.
Available at http://www.raritan.com/power-distribution-and-capacity. Accessed May 22, 2014.
Raritan Inc., “Power Distribution Units (PDUs): Power Monitoring and Environmental Monitoring to Improve
Uptime and Capacity Planning,”. White Paper,. Somerset (NJ): Raritan Inc., Somerset, NJ,; 2009c.
Available at http://www.raritan.com/power-and-environmental-monitoring. Accessed May 22, 2014.
Raritan Inc., “Deploying High Power to IT Equipment Racks,”. White Paper #V1156,. Somerset, (NJ): Raritan
Inc., Somerset, NJ, 2012.
Verdun, Gary. Editor. “The Green Grid Metrics: Data Center Infrastructure Efficiency (DCiE) Detailed Analysis,”.
White Paper #14,. Beaverton, (OR): The Green Grid, 2008.
UL 60950-1, “Information Technology Equipment – Safety – Part 1: General Requirements”, Underwriters
Laboratories, Inc. 2013.
To learn more about Raritan PDUs, visit: www.raritan.com/ipdus
Reference
77
Index
Administration and Maintenance, 45
Key Terms to Know, 73
Air Pressure Sensors, 40
Load Balancing, 51
Airflow Sensors, 39
Local and Remote Control/Switching, 45
Approval Agencies, 22
Local Display and User Interface, 35
Benefits of an Intelligent Rack PDU, 57
Multiple Power Supplies, 51
Branch Circuit Load Capacity, 14
Nameplate Data, 21
Branch Circuit Protection, 31
Open Point of Integration, 45
Branch Circuits, 14
Other Sensors, 40
Branch Circuits: Rated Current, 16
Outlet Density and Types , 30
Branch Circuits: Rated Voltage, 14
Overload Protection, 23
Circuit Breaker Metering, 33
Overview and Class of Devices, 11
Circuit Breakers: Single Pole vs. Double and
Triple Pole, 33
Physical Topology, 40
Communication Protocols, 40
Connectors: Ethernet, Serial, Sensor, USB and
Other, 31
Contact Closure Sensors, 40
Cord Length and Feed, 34
Cord Retention, 34
Customizing IT Equipment for Power Efficiency,
67
Data Collection, 42
Dual Feed to Dual Rack PDUs, 50
Dual Feed to Single Rack PDU with Transfer
Switch, 49
Electrical Power Distribution to the Rack, 14
Electrical Terminology, 13
Plugs, Outlets, and Cords,
Power Available and Distributed to Racks,
49
Power Efficiency, 58
Power Rating vs. Load Capacity, 21
Power Requirements of Equipment at
Rack, 53
Proper Grounding, 22
PUE Levels, 59
Rack PDU 208V Single-Phase vs. 208V
Three-Phase, 54
Rack PDU 400V Three-Phase, 55
Rack PDU Functionality, 56
Rack PDU Management System, 41
Element Management, 44
Rack PDU Selection and Special Application Requirements, 55
Elements of the System, 27
Rack PDU Selection, 55
Environmental Management, 37
Rack PDU, 29
Fault Management, 44
Ratings and Safety, 21
Form Factor, 30
References and Further Reading, 77
Fundamentals and Principles, 9
Remote User Interface, 35
Graphical User Interface (GUI), 43
Reporting and Analytics for Power Monitoring and Measurement, 43
Higher-Density, Higher-Power Rack PDUs with
Sensors, 66
Security of Data and User Access, 45
Humidity Sensors, 39
Single Feed to Single Rack PDU, 49
Increased Intelligence at the Rack to Support
Efficiency Initiatives, 67
Single-Phase or Three-Phase Input Power
for Rack PDUs , 29
Inrush Current, 53
System Connectivity, 40
Integration with Higher Level Data Center
Management Systems, 68
Temperature Sensors, 37
Introduction to Power Usage Effectiveness
(PUE), 58
The Advantages of High Power, 61
Types of Rack PDUs, 11
Why Advanced, Level 3 PUE?, 60
78
The iPDU Handbook
Understanding Rack PDUs with Ease
The Intelligent Power Distribution Unit (iPDU) Handbook is the definitive guide to rack PDUs
and understanding how they can change the face of your data center. Written with the rack
PDU novice in mind, this simple but elegant illustrated guide will cover the fundamental
principles of powering your IT equipment using single or three-phase power while leveraging
the advanced capabilities of different rack PDU models in order to increase uptime and availability while improving power efficiency.
Call 1.800.724.8090 or visit www.raritan.com