open rack charter - Open Compute Project

open rack charter - Open Compute Project
Open Rack Charter
Authors: Gio Coglitore, Pierluigi Sarti, John Kenevey, Pete Bratach
Contents Contents ....................................................................................................................................................... 2 Mission Statement .................................................................................................................................... 3 Summary....................................................................................................................................................... 3 Overview ....................................................................................................................................................... 4 License ...................................................................................................................................................... 4 Design Approach ........................................................................................................................................ 5 Physical Design ...................................................................................................................................... 5 Power Distribution (Bus Bars) ........................................................................................................... 6 Battery Backup Power ......................................................................................................................... 8 Cable Management ............................................................................................................................... 8 Compute Density and Thermal Design ........................................................................................... 8 Serviceability .......................................................................................................................................... 9 Installation Flexibility ......................................................................................................................... 9 Component Upgrade ............................................................................................................................ 9 Roadmap ....................................................................................................................................................... 9 Community Organization and Meeting Cadence ..........................................................................10 2
January 18, 2012
Open Rack Charter
Mission Statement Reducing TCO, improving component refresh cycles and increasing flexibility in the
scale compute space through simple, innovative and modular designs; an open form
factor that still allows for value-added differentiation at the component layer; a
standardized rack infrastructure with common interconnects and power distribution;
and optimized supply chain management.
Summary The Open Compute Project's focus is on server, storage and data center design. The
Open Rack design is a foundational technology for other Open Compute designs that
aims to diminish or eliminate the traditional differentiation at the mechanical rack and
power distribution layer. Open Rack is an open design that focuses on enabling
standard mechanical form factors, modular power supplies, improved cable
management, efficient thermal design, and flexible serviceability allowing fluid
component upgrades and independent refresh cycles.
The intent is to enable what we term component disaggregation. Disaggregated
compute components can be swapped and expire according to their own life cycles
without replacing the entire server for an upgrade or a repair. The open form factor
offers component choice without locking in to a specific vendor. This improves total
cost of ownership and allows needed flexibility in the scale compute space.
Typical server com ponent life cycles vs. the typical server replacem ent tim e.
http://opencompute.org
3
Overview The Open Rack can be thought of in the following areas:
•
•
•
•
•
•
•
•
Mechanical design
Power distribution
Battery backup power
Cable management
Compute density and thermal design
Serviceability
Installation flexibility
Component upgrades
The problems the Open Rack seeks to solve encompass:
•
•
•
•
•
•
•
Different mechanical form factors require different implementation schemes
and force vendor lock-in without associated benefit, as typified by blade servers.
Redundant power distribution integrated into and specific to each server.
Outdated cable management with proprietary interconnects.
Impractical compute density, restricted by thermal design.
Inefficient serviceability.
Inflexible installation processes.
Inefficient component upgrade cycles.
License As of April 7, 2011, the following persons or entities have made this Specification
available under the Open Web Foundation Final Specification Agreement (OWFa 1.0),
which is available at
http://www.openwebfoundation.org/legal/the-owf-1-0-agreements/owfa-1-0
Facebook, Inc.
You can review the signed copies of the Open Web Foundation Agreement Version 1.0
for this Specification at http://opencompute.org/licensing/, which may also include
additional parties to those listed above.
Your use of this Specification may be subject to other third party rights. THIS
SPECIFICATION IS PROVIDED "AS IS." The contributors expressly disclaim any warranties
(express, implied, or otherwise), including implied warranties of merchantability, noninfringement, fitness for a particular purpose, or title, related to the Specification. The
entire risk as to implementing or otherwise using the Specification is assumed by the
Specification implementer and user. IN NO EVENT WILL ANY PARTY BE LIABLE TO ANY
OTHER PARTY FOR LOST PROFITS OR ANY FORM OF INDIRECT, SPECIAL, INCIDENTAL, OR
CONSEQUENTIAL DAMAGES OF ANY CHARACTER FROM ANY CAUSES OF ACTION OF ANY
KIND WITH RESPECT TO THIS SPECIFICATION OR ITS GOVERNING AGREEMENT, WHETHER
BASED ON BREACH OF CONTRACT, TORT (INCLUDING NEGLIGENCE), OR OTHERWISE, AND
4
January 18, 2012
Open Rack Charter
WHETHER OR NOT THE OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
DAMAGE.
Design Approach Using the mission statement as a guide, our approach to deliver and maintain projects
within this track is:
•
•
•
•
•
•
•
•
Any solution must scale to 100,000+ physical machines.
Use existing standards whenever possible. If standards require adjustment,
partner with standards bodies to make changes.
Define requisite features, attributes and interface standards for OCP hardware
designs.
Encourage and work with the supplier ecosystem across the technology stacks
that Open Rack relies upon.
Maintain interoperability for equipment used in the Open Rack.
Work with OCP hardware designers to implement required functionality.
As part of every deliverable, a methodology to validate the functionality and
maintain validity thereof must be included.
As it is common to have some non-scale platforms in a scale environment,
encourage specialty hardware manufacturers to comply with the Open Rack
specification.
Physical Design The Open Rack design should remove unnecessary differentiation at the mechanical
design level, eliminating customer lock-in to a particular supplier. Customers and
suppliers alike can configure their own designs for each power zone.
The Open Rack form factor features:
•
•
•
•
•
•
23.6" wide rack, as floor steps in most data centers are 24". This provides
flexibility for the chassis form factor within a 21" wide equipment bay. Racks are
traditionally 19" wide.
"Practical density." We have designed a larger aperture (the opening in the front)
to improve airflow and create a good thermal profile. This allows for higher
compute density.
36-48" rack depth makes serviceability more practical and thermal flows more
efficient. Normal equipment depth is 33", with the remainder used for bus bars
and cables.
Up to 48U rack height maximize the density allowed while also allowing
deployment into most data centers. The accepted rack height is 95".
At least 1 power zone per column. A power zone comprises the power shelf
solution and the equipment bay that it powers. Power solution can be singlephase AC or three-phase AC, or HVDC. The width of the power shelf can be as
wide as the equipment bay itself. The height is measured in 0.5U increments.
The Open Rack increases compute density, demonstrates flexibility in design,
improving efficiency and lowering cost.
http://opencompute.org
5
•
•
•
•
•
Maximum of three 1U rack switches, allowing iteration on the compute density
reinforced by one or more power zones.
Compatible with all form factors and provides interoperability between
different suppliers.
A modular chassis that accommodates multiple sleds, so there can be multiple
motherboards in a given chassis in either horizontal or vertical layouts; the
chassis can be of varying height and can be pre-integrated into the rack.
A universally flexible combination mounting scheme for the equipment bay that
provides for fixed or adjustable mounting options in a variety of configurations
in 0.5U increments.
A slot in the equipment bay is measured in 0.5OU steps. An OpenU (OU) is 48mm.
Power Distribution (Bus Bars) Traditionally, power distribution is not modular but specific to a particular compute
system. The Open Rack's power is not part of the server unit but is designed into the
rack, putting the focus on efficiency.
The power distribution design enables redundant power schemes and also distributes
more power to a given device than a single PSU. This provides flexibility in compute
designs and flexibility to reuse the PSU and adjust the density profile.
The bus bars are designed for safety, efficiency and tolerance. They incorporate these
features:
•
•
•
•
•
•
•
•
•
•
6
Simplicity of design
Improved airflow
Seamless mating of power solution and compute unit
Improved efficiency for 12V distribution via multiple bus bars and multiple
power zones
The 12V "secondary" voltage provided by the bus bar is below the 60V SELV
threshold (see
http://www.electropedia.org/iev/iev.nsf/display?openform&ievref=851-15-08 and
http://en.wikipedia.org/wiki/Extra-low_voltage) and allows the design to bypass
regulatory oversight (which begins at 90V)
Very tight (constrained) mechanical tolerances enable mating with the compute
units
Up to 3 bus bar pairs provide ability to install 3 motherboards per slot
Eliminates bulky power cords that hinder airflow and are another point of failure
themselves
Enables redundancy without significant added CapEx
Optimizes efficiency of power distribution through higher utilization
January 18, 2012
Open Rack Charter
Overhead view of bus bars with vertical PD U.
http://opencompute.org
7
Power shelf connection to vertical bus bars (can be either cables or blades). Exam ple
shows single-phase power solution.
Battery Backup Power The Open Rack can use a flexible battery backup strategy in lieu of a centralized UPS. It
may be:
•
•
•
BBU within the rack or
OCP battery cabinet as a local backup source
48Vdc from a central backup line or HVDC scheme
Cable Management In consideration of cable management, there are some next generation interconnects
that make it better in the data center from management, interconnect-agnostic and
bandwidth points of view.
Compute Density and Thermal Design The Open Rack enables compute density that takes into account a practical thermal
design and efficiency. It is not focused on aperture reduction (that is, maximizing the
airflow over each server). Its form factor allows for large number of systems in the rack
without cost of thermal restrictions.
8
January 18, 2012
Open Rack Charter
Serviceability The Open Rack should be tool-less with easy-to-remove components that require the
least amount of effort to remove.
Installation Flexibility The rack can be delivered fully integrated and needing only to be bolted into the data
center. Or the rack can be ready to receive compute components during an
infrastructure build-out at the appropriate time (and cost), following a data center's
build-to-deliver cycle.
Component Upgrade The Open Rack modular design enables independent refresh cycles for all components.
V1 of this specification enables independent server and storage device upgrades
without modifying the power distribution scheme. We envision v2 of the spec to enable
disaggregation of other components such as network interface, DRAM, and boot
devices, as well as include a BBU solution in the rack.
Roadmap Ownership of non-recurring engineering (NRE) funds, manufacturing, validation and
test plans will vary based on market demand and supplier interest.
Version
Notes
Validation
1.0
Scheduled for 4Q2012, it features a 4.2KW
power shelf for an OCP server with an OCP
Intel v2.0 motherboard.
This mid-point release, scheduled for late 2012,
features greater than 5KW per power shelf,
wide-range rectifier schemes that can take an
estimated 120 to 240VDC, 1U with 1U lithium
polymer battery underneath, and enables HPC
and other markets to meet their business
requirements (such as better cost structures).
Planned for 2013. Focus is on superior design
for greater efficiency in the scale compute
space and other features. The form factor can
accommodate a high-voltage DC feed, still
with delivery to low voltage DC distribution
system within the rack.
Facebook will perform a
full evaluation for
production readiness.
The community's
evaluation is welcome.
1.5
2
http://opencompute.org
Facebook will perform a
full evaluation for
production readiness.
9
Community Organization and Meeting Cadence The Open Rack Project Committee chair is Gio Coglitore, Facebook. The committee also
includes:
•
•
•
•
Andy Bechtolsheim, Arista Networks
Stephen Chan, Facebook
Shesha Krishnapura, Intel
Pierluigi Sarti, Facebook
The committee operates under these rules:
•
•
•
10
The project committee chair runs regular meetings and ensures minutes are
recorded and shared with the community via the mailing alias. (The chair can
elect a secretary from the project committee to record minutes.)
Oral contributions fall under the OWF CLA and OWFa definition (see
http://www.openwebfoundation.org/legal/the-owf-1-0-agreements/owfcontributor-license-agreement-1-0---copyright-and-patent) and can equate to
actual contributions. They must be put in writing in order to be considered as a
contribution to a specification.
Minutes should be published to the project mailing alias 10 days after the
meeting.
January 18, 2012
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising