Académique Documents
Professionnel Documents
Culture Documents
Navindra Yadav (Principal Engineer), nyadav@cisco.com Eruch Kapadia (Solution Architect), ekapadia@cisco.com December 2, 2010
2010 Cisco Systems, Inc. All rights reserved. -Ver 0.3
Grid-Interop 2010
1
Outline
Reliability and Availability Basics Inter Substation Traffic
Use Cases like Tele Protection, Primary Protection, etc Reliability and Design improvement
Grid-Interop 2010
2
Grid-Interop 2010
3
Measurement of Availability
Availability = (MTBF)/(MTBF + MTTR)
MTBF = Mean Time Between Failure MTTR = Mean Time To Repair
Grid-Interop 2010
4
9 11
10
12
13
14
15
Failure of any node, link disrupts Network Availability = A = A1 * A2 * A3 * A4 * A5 * A6 * A7 If all A* are 90%. A = 0.4782969 = 47%
Assume a simple Active-Active Design Network Availability = 1 (UnAvail_P1 * UnAvail_P2 * UnAvail_P3) = = 1 ((1-Avail_P1) * (1-Avail_P1) * (1Avail_P1)) If all A* are 90% A= 1 ((1-0.59049)* (1-0.729) * (10.4782969)) = 1 (0.40951* .271 * . 0.5217031) = 0.942102845513649 = 94%
5
Grid-Interop 2010
IEC 61850 GOOSE and SV over the WAN Challenges and Solutions
Grid-Interop 2010
6
Gateway
Example Proxy Gateways GWs Must Terminate Protocols GWs must Understand Applications and configuration changes in the application Latency and Jitter addition, especially when GWs are implemented in software
Grid-Interop 2010
7
Problem: IEC 61850 GOOSE/SV over WAN Layer2 Tunneling Fault Domain Extension Lower Availability
This approach creates Large Fault Domains
Substation network Faults spread Dramatically lower availability Faults in unrelated parts of the network propagate Calculation of Availability means factoring Availability of the entire L2 domain means UNRELATED networks too!!
Grid-Interop 2010
8
Grid-Interop 2010
9
Problem: Layer 2 GOOSE / SV over the WAN Implications on Scaling, Security, Replication, Flooding, etc
Issues:
Intra Substation Replication Inter Substation Replication Information Leakage Security Implications Wasted Bandwidth Limited Scale
2010 Cisco Systems, Inc. All rights reserved.
Grid-Interop 2010
10
Solution: IEC 61850 with IPv4/v6 profile provides Scalability, Security, etc
GOOSE/SV on IPv4/v6 routable protocol
Scalable Low (in usecs) Latency All HW forwarding Path Low (in usecs) Jitter Cyber Security benefits Easy to trouble shoot and manage over WAN proven model
Grid-Interop 2010
11
IEC 61850-90-5 for PMUs is working on a 61850 profile to carry GOOSE / SV over TCP/IP[v4v6] stack IP profiles being developed above must Not only be restricted to PMUs, but also to other all relays and applications.
Grid-Interop 2010
12
IED/Relay Vendor Support Scalability Tunnels / Encapsulation WAN Availability / Reliability Inherent Network State
Inter Substation Traffic Replication (Multicast) Intra Substation Traffic Replication (Multicast)
2010 Cisco Systems, Inc. All rights reserved.
Efficient. WAN Network replicates at most optimal points Efficient. No flooding inside the substation multicast delivered to only interested hosts (IGMP/MLD snooping)
Grid-Interop 2010
13
Comparison Contd
Topic
Latency
Higher Jitter (order of multiple msecs) Weaker. Large Flooding domain across substations create security challenges (like spoofing, replay, DOS, info leak domains)
State Scaling Domain (Eg. Mac Table size) Layer 2 Media dependent solution
Inferior. Limited all substations learn of everyone else thanks to flooding Yes
Superior. Only stations that communicate with each other need to know about each other No
Grid-Interop 2010
14
Grid-Interop 2010
15
Background
A lot of discussions going on Ethernet topologies for Substation Two Leading topologies
Rings Redundant Trees
The Utility industry is familiar with Rings Cisco has massive deployment expertise with
Ring Topologies Redundant Tree Topologies and many others
There are applications where both topologies make sense Cisco wants to share its experience with these different Topology choices Next few slides compare Pros/Cons of Rings Vs Redundant Tree topologies
Majority of Ethernet deployed world wide in IT space (Enterprise/Datacenter) is based on Redundant Tree Topologies
Evaluate these Topologies (and their Pros/Cons) when you are designing
New Substation Designs Substations going through major retrofits
Grid-Interop 2010
16
Grid-Interop 2010
17
Grid-Interop 2010
18
Redundant Trees
Yes More conductor. If conductors are laid inside trenches and conduits with limited capacity for extra cables its an issue for trees. Fixed and deterministic latency. Tree depth determines the number of hops. Smaller Fault Domain (fault limited to the triangle of switches in the tree) All inter switch traffic contends at limited and few fixed points in the tree topology. Superior (only the the root switches need to have policies and mac addresses for every device. The leaf switches just need to have capacities to support their downstream end devices) Superior (no downtime to the network to add a new leaf switch) Fewer and equal number of contention points, through which all traffic passes when going between two access switches, yields a fairer system Fast convergence can be achieved by using FlexLinks. Zero down time with PRP (from IEC 62439). Fewer and fixed number of switches in the switching path results in a higher MTBF. As there are fewer switches to switch through. Also aggregation switches can be designed for higher availability. Superior (not all switches have to have all vlans, also the flooding domains are smaller) Authenticator Function (for protocols like 802.1x) only on Switches 2010 Cisco Systems, Inc. All rights reserved.
Rings
Yes Simpler, Less conductor/fiber. In some cases trench and conduit size may make ring the only viable option. Latency varies. The number of hops between the source and the destination depends on where the loop in the ring is broken. When the blocking point changes the latency also changes. The whole ring is the fault domain. All inter switch traffic contends for the ring bandwidth
Predictable Latency
Inferior (all switches have to learn about all end points. Least capable switch determines the capacity of the ring)
Inferior (downtime seen by the ring when a switch is added or deleted from the ring) Traffic sent by the edge switches has to compete with similar class of traffic at every hop on the ring, the contention points and their number can change over time. Also the contention points vary between two access switches. Faster convergence (sub 50 msecs) can be achieved by using some like REP. Zero down time with PRP (IEC 62439) protocols. Variable MTBF as the number of switches in the number of switches vary depending on the topology.
Inferior. (all switches have to have all vlans). If the Ring nodes are end points then over all security suffers. End hosts do not authenticate other end hosts
Authenticator Function (for protocols like 802.1x) has to be pushed to all 19 devices on the ring like end devices
Grid-Interop 2010
Grid-Interop 2010
20
Designs to meet or exceed 6+ Nines of Availability (Proactive and Reactive Redundancy) with Redundant 61850 Actors
WAN
WAN
SUBSTATION
Blue Network is Active Primary Network. All devices are at least attached to the Blue network
Green Network is Active Redundant Network. Smaller in size as not all devices will be dual attached
Other
MU
BPU
BIED
MU
BPU
BIED
Simplified view of the Power Network
Power Line
Superior Availability Characteristics Above can be built using a Tree or Ring topologies Active-Active Design No down time due to network convergence events Zero Down time for any failure N-1 Redundancy for any failure Can be improved even further if Blue and Green end points can process each others updates besides their own too. Get to N-2 redundancy
2010 Cisco Systems, Inc. All rights reserved.
Grid-Interop 2010
21
Designs to meet or exceed 6+ Nines of Availability (Proactive and Reactive Redundancy) without Redundant Actors
WAN
WAN
SUBSTATION
Blue Network is Active Primary Network. All devices are at least attached to the Blue network
Green Network is Active Redundant Network. Smaller in size as not all devices will be dual attached Using say IEC 62439 - PRP
Other
MU
BPU
BIED
Power Line
Superior Availability Characteristics Above can be built using a Ring topology too Active-Active Design No down time due to network convergence events Zero Down time for any network device failure
2010 Cisco Systems, Inc. All rights reserved.
Grid-Interop 2010
22
Navindra Yadav
nyadav@cisco.com
Eruch Kapadia
ekapadia@cisco.com
Grid-Interop 2010
23