1. Ethernet Automatic Protection Switching (EAPS)
A small comparison with Eternet Ring Protection Switching (ERPS)

2. Introduction
EAPS is a protocol invented to increase the availability of Ethernet rings
Developed by Extreme Networks (RFC3619 – 2003)
Objective:
Provide a resilience level comparable to SONET rings
Current version (v ) has some enhancements over version 1 (RFC3619 – 2003)

3. Motivation
Ethernet is widely used in Local Area Networks (LANs) and Metropolitan Area Networks (MANs)
Typically present a ring topology
MAN operators want to reduce recovery time
Spanning Tree Protocol (STP) could take 30 – 60 second to recover
Rapid Spanning Tree Protocol (RSTP) is faster...
Convergence time depends on the number of nodes
Both STP and RSTP limit the number of nodes
EAPS recovers in less than 1 second (100 ms)
Does not limit the number of nodes!!!

4. Basic Considerations (I)
A ring is made up of two or more switches
Each switch has two ports connected to the ring
An EAPS domain exists on a single Ethernet ring
A domain protects a group of VLANs
A domain has a unique control VLAN
Multiple EAPS domains could coexist on the same ring
Multiple control VLANs

5. Basic Considerations (II)
For each EAPS domain:
One of the nodes is the Master (S1)
One port is designated as the Primary port (P)
The other is the Secondary Port (S)
All other nodes (S2-S6) are known as Transit nodes

6. Normal Operation
The Master node blocks its secondary port -> avoid loops
Non-control traffic is blocked (Control VLAN is NOT blocked)
Master is in COMPLETE state
Transient nodes are in LINKS-UP state
The Master sends health-check frames (HEALTH-CHECK- PDU) periodically (Hello timer)
From primary port to secondary port
Control frames consumed by the Master -> NOT forwarded

7. Fault Operation When a fault is detected:
The Master changes to FAILED state
Unblocks secondary port
Flushes it bridging table
The Master orders the other nodes to flush their tables
Sends a RING-DOWN-FLUSH-FDB-PDU frame
Transit nodes learn the new topology

8. Fault Detection (I) 2 ways of detecting a failure Link Down Alert
Ring Polling
Transient nodes detect a link-down
Transient detecting the failure changes to LINKS-DOWN state
Transient sends a LINK-DOWN-PDU frame to the Master
Master changes to FAILED state
Master unblocks secondary port
...

9. Fault Detection (II) Ring Polling (version 1 – RFC3619)
Master sends HEALTH-CHECK-PDU frames periodically
From primary to secondary port
Master has a Fail-period timer
If health check frame received before timer expires -> reset timer
If health check frame NOT received before timer expires
Master changes to FAILED state
Master unblocks secondary port
...

10. Fault Detection (III) Ring Polling (version 1.3)
2 options if the Fail-period timer expires (configurable)
«Open Secondary Port» -> previous slice
«Send-Alert»
Master DO NOT unblock its secondary port yet
Master sends a QUERY-LINK-STATUS-PDU frame out of both ports
Transit nodes with link failure reply with LINK-DOWN-PDU frame
Master changes to FAILED state
...
Prevents False Failures
Health frames could not return to Master –> even if the ring is complete
Control VLAN misconfigurations
Too much traffic
Master node’s CPU busy
Why?

11. Fault Restoration (I)
Master in FAILED state -> continues sendind HEALTH-CHECK-PDU frames
Ring restored -> Master’s secondary port receives health frame
Master changes to COMPLETE state
Blocks non-control frames on secondary port
Flushes its bridge table
Orders the other nodes to flush their tables
Sends a RING-UP-FLUSH-FDB-PDU frame
Transit nodes re-learn the topology

12. Fault Restoration (II) – PREFORWARDING State
Time between
The Transit node detecting its link is restored
The Master detecting the ring is restored
Master’s secondary port is unblocked
Possible temporary loop !!!!
When Transit node detects its link is restored
Changes to PREFORWARDING state and starts Preforwarding timer
Protected VLANs in that port are temporary blocked
Waits till a RING-UP-FLUSH-FDB-PDU is received
Changes to LINKS-UP state
Unblocks previously blocked VLANs
Flushes its bridge table and stops Preforwarding timer
Re-learns topology

13. Fault Restoration (III) – PREFORWARDING State
Preforwarding timer deals with:
Lost RING-UP-FLUSH-FDB-PDU from the Master
Another break in the ring
If the transient node remains in PREFORWARDING state indefinitely -> disconnected network
Preforwarding timer is derived from the Hello-timer for HEALTH-CHECK-PDU frames

14. Enhancements of version 1.3
«Send-alert» configuration for Ring Polling fault detection method
INIT state
Master comes up for first time and its ports are up
Master does not know if the ring is up
Master starts in INIT state -> blocks secondary port
When the first health frame is received -> changes to COMPLETE state
Helps spotting misconfigurations in control VLAN
LINK-UP-PDU
Transient detects a link comes up -> sends LINK-UP-PDU to Master
Timestamp used for trouble-shooting
If the Master never changes to COMPLETE state
Allows use of EAPS Shared-Ports

15. VLANs in Multiple EAPS domains (Multiple Rings) (I)
EAPS could handle a simple configuration
Each ring has a EAPS domain, a Master node and a Control VLAN
VLAN spanning in both rings is added as protected by both EAPS domains

16. VLANs in Multiple EAPS domains (Multiple Rings) (I)
Topologies with a common link could be problematic
If the common link fails
Both Masters open secondary ports
Protected VLANs spanning both rings will have a loop
S1-S2-S3-S4-S5-S6-S7-S8-S9-S10-S1
EAPS Shared-Ports deals with it
Out of the scope

17. States and Control Frames
Version 1 – RFC3619
Version 1.3

18. Ethernet Ring Protection Switching (I)
Ethernet Ring Protection Switching (ERPS) is defined by ITU-T G > achieve sub-50 ms recovery times in rings
Basic considerations:
One link is designated as the Ring Protection Link (RPL) -> blocked to prevent loops
The node setting the block is the RPL Owner (Master in EAPS)
Nodes monitor link failure using Ethernet Continuinity Check (ETH-CC) messages
Four defined local events:
Local Signal Failure (local SF) -> detection of link failure
Local clear Signal Failure (local clear SF) -> detection of link restoration
Wait-To-Restore Expire (WTR-Expire) -> timer expiration
Wait-To-Restore Running (WTR-Running) -> timer running

19. Ethernet Ring Protection Switching (II)
Basic considerations (cont.):
The protocol uses Ring Automatic Protection Switching (R-APS) messages:
R-APS(SF): sent by the node detecting link failure (gets local SF)
R-APS(NR): sent by the node detecting link restoration (gets local clear SF)
R-APS(NR,RB): sent by RPL Owner indicating the RPL is blocked
Two important timers
Wait-To-Restore (WTR) Timer: used by the RPL Owner to verify that the ring has stabilized before blocking the RPL after failure
Guard Timer: used by links detecting link restoration to avoid receiving outdated R-APS messages
Three states for nodes
Initialization: first defining the node
Idle: normal state, RPL blocked, all nodes/ports working
Protecting: protection switching is in effect

20. Ethernet Ring Protection Switching (III)
Basic considerations (cont.):
An R-APS channel is configured using a VLAN -> transmitting R-APS messages

21. ERPS Principle of Operation (I)
In normal operation (nodes in state Idle): RPL is blocked
Link failure (local SF): nodes detecting it block failed port, send R-APS(SF) and flush filtering database (FDB)
Nodes receiving R-APS(SF) flush FDBs
RPL Owner receives R-APS(SF): flushes FDB, unblocks RPL
Link Restoration (local clear SF): detecting nodes send R-APS(NR) periodically and start Guard Timer
RPL Owner receives R-APS(NR): starts WTR Timer
WTR Timer expires: RPL Owner blocks RPL, sends R-APS(NR,RB) and flushes DFB
Nodes receiving R-APS(NR,RB) flush FDBs
Nodes detecting link restoration unblock recovered ports, stop sending R-APS(NR) and flush FDBs

22. ERPS Principle of Operation (II)

23. EAPS vs. ERPS
Same basic idea: break the loop in the ring by blocking one port
In case of failure, unblock the blocked port and keep connectivity
EAPS:
Both the Master and Transient nodes can detect a failure
Only the Master detects the failed link is restored
ERPS:
Only the nodes adjacent to a failed link detect failures and restoration

24. References
S.Shah, M. Yip, «RFC3619: Extreme Networks’ Ethernet Protection Switching (EAPS), Version 1», Network Working Group, October 2003.
A. Lim, S. Blake, S. Shah, «Extreme Networks’ Ethernet Protection Switching (EAPS), Version 1.3», Internet-Draft, July 2011.
Extreme Networks Whitepaper «Ethernet Automatic Protection Switching (EAPS)», Extreme Networks, Inc., 2006.
J. D. Ryoo, H. Long, Y. Yang, M. Holness. Z. Ahmad, J. K. Rhee, «Ethernet Ring Protection for Carrier Ethernet Networks», IEEE Comm. Magazine, September 2008