Software-Defined Networking Changes the Paradigm for

Software-Defined Networking Changes the Paradigm for
Software-Defined Networking
Changes the Paradigm for
Mission-Critical Operational
Technology Networks
Paul Robertson
© 2017 by Schweitzer Engineering Laboratories, Inc. All rights reserved.
All brand or product names appearing in this document are the trademark or registered trademark
of their respective holders. No SEL trademarks may be used without written permission. SEL
products appearing in this document may be covered by US and Foreign patents. 20170125
Introduction
While software-defined networking (SDN) was developed primarily to solve an information
technology (IT) problem by providing a more efficient approach to managing dynamic high-traffic
networks, SDN also offers significant advantages for operational technology (OT) networks.
While both OT and IT networks manage the transfer of data, OT networks support processoriented applications. They manage the process and automation systems that control power
systems, industrial plants, transportation infrastructure, and other critical systems. Unlike IT
networks that deal primarily with the movement of data, OT networks are focused on the real-time
decision making necessary for the operation of mission-critical systems. They carry information
from sensors, transducers, intelligent electronic devices, and automation controllers.
Ethernet has become an established technology for data networking, based largely on its success
in creating a common interface and protocols to enable interoperable communications between
devices and applications from multiple manufacturers. Ethernet’s convenience has led to its
widespread use in OT networks. However, for managing real-time processes, traditional Ethernet
has several limitations: slow healing times, forced behavior, no inherent security, difficult testing,
and no network visibility.
This white paper originally appeared as an article in [1].
SDN Explained
SDN is a new approach for managing the forwarding of Ethernet packets through a network. It is
revolutionizing the management and operation of large-scale IT enterprise networks, cloud
infrastructures, and data center systems. SDN provides better support for the highly dynamic
characteristics of these networks. SDN is ideally suited for OT networking as well because it
resolves the limitations of applying Ethernet to real-time OT systems.
Before explaining how SDN addresses the unique challenges of OT systems and how OT-focused
SDN offers a different value proposition than IT-focused SDN solutions, it is helpful to review the
technology. A basic SDN architecture is shown in Figure 1.
Application Layer
Operation, Administration, and
Maintenance Applications
Network Visualization
Control Plane
Configuration Programming
Network Operating System
OpenFlow®
Data Plane
Simple
Packet-Forwarding
Hardware
Simple
Packet-Forwarding
Hardware
Simple
Packet-Forwarding
Hardware
Figure 1
SDN Architecture
In SDN, the control plane—which determines what to do with each Ethernet packet by making
forwarding decisions—is separated from the data plane, which physically forwards the Ethernet
data packets.
3
In traditional Ethernet, the control plane is located in network appliance hardware (i.e., switches).
In SDN, however, the control plane is located in software called a flow controller. This allows the
switching hardware to be simpler and optimized for packet forwarding. At the same time, the flow
controller is able to take advantage of a software-oriented architecture to program and configure
all of the switches in the network. The combination provides a programmatic approach to
implementing networking functions that results in enormous flexibility, greater speed, and
increased security.
Several protocols have been defined to provide an interface between the control plane and the
data plane. OpenFlow, for example, is an interface specification developed by the Open
Networking Foundation to promote the adoption of SDN and enable interoperability between
different flow controller implementations and different manufacturer’s switch hardware.
In addition to the control plane, SDN defines an application layer that provides a method for
supporting operation, administration, and maintenance tasks. The application layer provides a
mechanism to extract meaning from and interact with the complexity of the low-level programming
rules and actions that control how each packet is forwarded. This simplifies the network
configuration process and enables users to configure their networks using a higher level of
abstraction.
How SDN Works
SDN uses the following terminology:
•
Match criteria specify Ethernet frame header information attributes that each ingress
packet is compared against.
•
An instruction defines a specific action or set of actions that is applied to all packets that
meet specified match criteria. Instructions define which port to forward a packet out of and
what path the packet takes across the network.
•
A flow is a group of attributes that defines a single communications session. Each flow is
defined by a set of match criteria and corresponding instructions that are applied to each
packet that ingresses a switch.
•
A flow table contains all of the flows defined within a switch.
•
The flow controller is the centralized controller that programs the rules and flow tables in
each switch.
The flow controller gives each switch a set of rules containing match criteria and instructions to
determine the operations that should be performed on each packet the switch receives. These
rules are stored in a flow table. The switch matches the packet header information against each
rule contained in the flow table. If a matching flow table entry is found for a packet, the actions
associated with that flow table entry are executed, such as sending the packet out a specific port
number, modifying a field, or dropping the packet.
Instructions can be predefined for a situation in which an SDN switch receives a packet it has
never seen before (i.e., for which it has no matching flow table entries). One very powerful attribute
of SDN is that it allows a deny-by-default security profile that can instruct each switch to drop
every packet that does not meet a predefined rule. Counters are maintained for each flow table
entry and used to maintain statistics on traffic by flow, port, queue, and so on.
Each switch has the ability to look deep into an Ethernet frame and use that information to
determine a rule and an associated action. This makes SDN switches true multilayer devices that
can operate across Layer 1 to Layer 4 of the Ethernet frame (see Figure 2).
4
SDN Flow Match Rule
Ingress Port
Layer 1
Ethernet Header
IP Header
TCP/UDP Header
Layer 2
Layer 3
Layer 4
Payload
Ethernet Frame
Figure 2
Multilayer Match Rules Forward Approved Packets
Layer 1 is the physical layer that defines the physical port and media that data are received and
transmitted over.
Layer 2 is the data link layer that establishes and terminates connections between physically
connected devices. It uses media access control (MAC) addresses to uniquely identify each
device.
Layer 3 is the network layer. It provides the mechanism to transfer variable-length data from one
node to another. It is responsible for packet forwarding between routers using Internet Protocol
(IP) address information.
Layer 4 is the transport layer. It is responsible for establishing connectivity between applications
on host devices. It uses sockets to define the end points in the network. A socket address is a
combination of the IP address and the port number. The port number defines the communications
protocol and associated application. Transmission Control Protocol (TCP) and User Datagram
Protocol (UDP) are the most commonly used protocols. In the following example, TCP Port 20000
is used, which defines Distributed Network Protocol (DNP3) for supervisory control and data
acquisition (SCADA) applications.
Because each SDN switch can operate as a multilayer device, it is possible to create complex
match and action rules that are application-aware. In the example shown in Figure 3, the switch
has been configured to only forward SCADA traffic between two specific devices. A rule has been
created to only forward DNP3 packets on TCP Port 20000 that are received through Port 1 and
meet the following criteria:
•
Source MAC address: 00:30:A7:06:29:01.
•
Destination MAC address: 00:30:A7:06:13:29.
•
Source IP address: 1.1.1.2.
•
Destination IP address: 2.2.2.2.
•
Destination TCP port: 20000.
If the packet matches these criteria, the switch forwards the packet out Port 3.
Physical
Port ID
1
Source
MAC Address
Destination
MAC Address
EtherType VLAN ID
00:30:A7:06:29:01 00:30:A7:06:13:29
*
*
Action
Packet
1
Output Forward Port 3
Figure 3
IPv4
IPv4
TCP/UDP TCP/UDP
Source Destination Source Destination
1.1.1.2
OpenFlow
Switch
2.2.2.2
2
3
4
OpenFlow Match and Action Example
5
*
Packet
20000
This multilayer rule example demonstrates several powerful concepts. SDN allows applicationspecific network paths to be engineered; it enables each network device to operate as a Layer 2,
3, and 4 device; and it supports the ability of each switch to operate as a deny-by-default gateway.
Applying SDN to OT
To date, SDN’s success has been attributed to its ability to solve challenges associated with
managing the large, dynamic IT networks with high and variable traffic loads primarily found in
data centers. These IT networks have changing usage patterns, unpredictable application
demand, and varying traffic volumes.
In contrast, OT networks are static. In an OT network, all applications and services are known at
the design stage, and a rule set can be predefined for each switch to support the application and
connectivity requirements of the network. Typically, very few changes are required to the network
over time, and traffic volume remains predictable.
Microsecond Failover Times
By optimizing switching hardware to efficiently execute the match rules within the flow tables, it is
possible to achieve microsecond-speed packet inspection, decision, and forwarding times. In
addition, by predetermining each primary and failover path, it is possible to perform path healing
in tens of microseconds, which is two orders of magnitude faster than the spanning-tree
algorithms (STAs) used by traditional Ethernet. Networks can heal quickly enough that the only
data loss is that of the packet being transferred during the link failure. In many cases, zero packets
are lost during failover events.
Predictable Behavior
With SDN, it is possible to define different forwarding paths for different applications. In the
example shown in Figure 4, engineering access, SCADA, and Generic Object-Oriented
Substation Event (GOOSE) traffic have each been given different forwarding paths across the
network. There are two paths defined for GOOSE, a primary path (GOOSE 1) and a failover path
(GOOSE 2). The GOOSE 2 path is on a dedicated port and cable that are separate from the
engineering access path.
Flow
Controller
PC
SDN Switch
IED
Combined
Traffic
SDN Switch
GOOSE 1
GOOSE 2
IED
SDN Switch
Engineering
Access
SCADA
SDN Switch
Figure 4
Control Packet Forwarding by Application
6
Automation
Controller
With SDN, it is possible to have multiple physical connections between the same switches to
provide physical redundancy and traffic segregation. This is not possible with traditional Ethernet
because STAs prevent multiple physical connections by blocking ports. By taking this traffic
engineering approach, predictable delivery and failover performance can be achieved, and
network performance can be validated under all contingencies prior to commissioning.
Stronger Network Security
Within an OT environment, new applications should not start communicating across the network
to meet changing end user needs. As a consequence, the network does not rely on constant
participation with the flow controller to manage updates to flow tables. After the initial configuration
of each switch by the flow controller, the network can be locked down with a rule set that only
allows predefined application traffic to be forwarded. This provides a significant security
advantage over traditional Ethernet. Each switch provides a deny-by-default function on every
incoming packet. If the packet does not match a predefined rule, it is dropped by default.
Alternatively, it can be forwarded to a dedicated Intrusion Detection System (IDS) for further
analysis. SDN offers significant network security advantages to help address concerns about the
vulnerability of critical infrastructure systems, such as the power grid, to cyberattacks.
Simplified Design and Test
One of the biggest challenges with designing and implementing Ethernet networks is verifying
that the network will behave as designed, particularly during failover scenarios. There is no easy
way to perform contingency testing other than pulling cables and using a protocol analyzer to
verify that the STA being used has converged as expected based on the root switch and network
topology.
SDN changes the paradigm for network design and testing. Using software to abstract the
implementation of each rule, the network designer can specify the path each application should
take through a network and predefine each failover path. The network can then be validated for
each contingency using a programmatic approach that ensures all failure scenarios are tested
and validated.
Interoperability
SDN is based on Ethernet and does not require the modification of the Ethernet packet in order
to be implemented, so it is interoperable with standard Ethernet. Care should be taken when
establishing the boundaries between SDN and traditional managed Ethernet networks that rely
on protocols to operate. Protocols, such as Border Gateway Protocol (BGP), Rapid Spanning
Tree Protocol (RSTP), and Open Shortest Path First (OSPF), are in constant conversation in a
traditional managed Ethernet network. SDN does not use these protocols to manage network
operation, and therefore, SDN cannot be mixed with managed Ethernet switches across a single
topology. Distinct boundaries between an SDN and a managed Ethernet network should be
created, and rules should be implemented to instruct an SDN switch on how to appropriately deal
with managed Ethernet protocol packets.
Network Visibility
Lack of network management capability has always been a limitation of traditional Ethernet. By
putting the control plane in a centralized entity, the flow controller can monitor the configuration,
health, and status of the complete network.
7
Conclusion
SDN offers a fundamental change to the way Ethernet networks determine how to forward
packets and to how they are managed by the network administrators. SDN removes the control
plane from the physical switching hardware and locates it in a centralized flow controller. By doing
so, switches change from being distributed, intelligent, autonomous devices to being simpler
entities that perform a set of instructions provided by the flow controller. A suite of protocols
communicating in the background is no longer needed to manage how each switch discovers
neighbors, prevents loops, and recovers from link failures.
The primary benefit of SDN is its ability to support rapid changes to network operation and
topology to address fast-changing user needs. Many view SDN primarily as a solution for dynamic
IT networks. However, the principles of SDN offer significant benefits to time-critical OT
networking by addressing the limitations that traditional managed Ethernet has in terms of slow
healing, forced behavior, lack of security, difficult testing, and lack of network visibility. Table 1
provides a summary of the key advantages that SDN provides over traditional Ethernet. SDN
offers a performance breakthrough in all of these areas and promises to redefine Ethernet for
mission-critical applications.
Table 1 Comparison of SDN and Ethernet Performance
Attribute
SDN
Traditional Ethernet
Healing
<0.1 ms
>10 ms
Predictability
Yes
No
Security
Deny-by-default
Based on trust
Design and testing
Simple
Difficult
Interoperability
Yes
Yes
Network visibility
Yes
No
Reference
[1]
P. Robertson, “Software-Defined Networking for Mission-Critical Operations,” Industrial
Ethernet Book, Issue 98, February 2017. Available: http://www.iebmedia.com.
Biography
Paul Robertson is a senior marketing program manager for the communications product lines at
Schweitzer Engineering Laboratories, Inc. (SEL). He has over 25 years of experience developing
and marketing products for the telecommunications industry, spanning cellular wireless and wire
line communications systems. Paul worked in various technical and marketing roles for Motorola,
Hewlett-Packard, and Agilent Technologies before joining SEL. He has a B.Eng. in electrical and
electronic engineering from the University of Strathclyde and an M.B.A. from Edinburgh Business
School.
Schweitzer Engineering Laboratories, Inc.
Tel: +1.509.332.1890 | Email: [email protected] | Web: www.selinc.com
8
*LWP0016-01*
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

advertisement