G.8032 Ethernet Ring Protection Overview. March, 2008 ITU-T Q9 SG 15

Transcription

1 G.80 Ethernet Ring Protection Overview March, 008 ITU-T Q9 SG 5

2 genda G.80 Recommendation Introduction G.80 Objectives and Principles G.80 Concepts G.80 Protection Switching G.80 R-PS Messages G.80 Items Under Study

3 ITU-T SG5/Q9 G.80 Recommendation G.80 Recommendation consented February 008 This Recommendation defines the PS protocol and protection switching mechanisms for ETH layer ring topologies. Included are details pertaining to bridged ring protection characteristics, architectures and the ring PS protocol. G.80 Proposals and Contributions G.80 Recommendation Consent G.80 Objective, Principles, and Requirements Setting G.80 Converged Proposals G.80 ERP Work Item pproved in ITU-T J N F E M R P R M Y J U N J U L U G S E P O C T N O V D E C J N F E M R P R M Y J U N J U L U G S E P O C T N O V D E C J N F E M R

4 G.80 Objectives and Principles Use of standard 80 MC and OM frames around the ring. Uses standard 80.Q (and amended Q bridges), but with xstp disabled. Ring nodes supports standard FD MC learning, forwarding, flush behaviour and port blocking/unblocking mechanisms. Prevents loops within the ring by blocking one of the links (either a pre-determined link or a failed link). Monitoring of the ETH layer for discovery and identification of Signal Failure () conditions. Protection and recovery switching within 50 ms for typical rings. Total communication for the protection mechanism should consume a very small percentage of total available bandwidth.

5 G.80 Terms and Concepts Ring Protection Link () Link designated by mechanism that is blocked during Idle state to prevent loop on ridged ring Owner Node connected to that blocks traffic on during Idle state and unblocks during Protected state Link Monitoring Links of ring are monitored using standard ETH CC OM messages (CFM) Signal Fail () Signal Fail is declared when ETH trail signal fail condition is detected No Request (NR) No Request is declared when there are no outstanding conditions (e.g.,, etc.) on the node Ring PS (R-PS) Messages Protocol messages defined in Y.7 and G.80 utomatic Protection Switching (PS) Channel - Ring-wide VLN used exclusively for transmission of OM messages including R-PS messages 5

6 G.80 Timers G.80 specifies the use of different timers to avoid race conditions and unnecessary switching operations WTR (Wait to Restore) Timer Used by the Owner to verify that the ring has stabilized before blocking the after Recovery Hold-off Timers Used by underlying ETH layer to filter out intermittent link faults Faults will only be reported to the ring protection mechanism if this timer expires 6

7 Controlling the Protection Mechanism Protection switching triggered by Detection/clearing of Signal Failure () by ETH CC OM Remote requests over R-PS channel (Y.7) Expiration of G.80 timers R-PS requests control the communication and states of the ring nodes Two basic R-PS messages specified - R-PS() and R-PS(NR) Owner may modify the R-PS(NR) indicating the is blocked: R-PS(NR,R) Ring nodes may be in one of two states Idle normal operation, no link/node faults detected in ring Protecting Protection switching in effect after identifying a signal fault 7

8 Ring Idle State. Physical topology has all nodes connected in a ring. ERP guarantees lack of loop by blocking the (link between 6 & in figure) C. Logical topology has all nodes connected without a loop. Owner D. Each link is monitored by its two adjacent nodes using ETH CC OM messages E. Signal Failure as defined in Y.7, is trigger to ring protection Loss of Continuity Server layer failure (e.g. Phy Link Down) 6 5 Physical topology Logical topology

9 Protection Switching Link Failure. Link/node failure is detected by the nodes adjacent to the failure.. The nodes adjacent to the failure, block the failed link and report this failure to the ring using R-PS () message C. R-PS () message triggers Owner unblocks the ll nodes perform FD flushing D. Ring is in protection state E. ll nodes remain connected in the logical topology. R-PS() 6 R-PS() R-PS() Owner Physical topology R-PS() Logical topology 5

10 Protection Switching Failure Recovery. When the failed link recovers, the traffic is kept blocked on the nodes adjacent to the recovered link. The nodes adjacent to the recovered link transmit R- PS(NR) message indicating they have no local request present C. When the Owner receives R- PS(NR) message it Starts WTR timer D. Once WTR timer expires, Owner blocks and transmits R-PS (NR, R) message E. Nodes receiving the message perform a FD Flush and unblock their previously blocked ports F. Ring is now returned to Idle state R-PS(NR) R-PS(NR) Physical topology Logical topology R-PS(NR, R) R-PS(NR) Owner R-PS(NR, R-PS(NR) R)

11 Signaling Channel Information ERP uses R-PS messages to manage and coordinate the protection switching R-PS defined in Y.7 - OM common fields are defined in Y.7. Version for this version of Recommendation OpCode defined to be 0 in Y.7 Flags should be ignored by ERP MEL Version (0) OpCode (R-PS = 0) Flags (0) TLV Offset () 5.. R-PS Specific Information ( octets) 7 [optional TLV starts here; otherwise End TLV] last End TLV (0) Defined by Y.7 Defined by G.80 Non-specified content

12 R-PS Specific Information Specific information (octets) defined by G.80 Request/Status(bits) 0 = 0000 = NR Other = Future Status R (bit) Set when is blocked (used by Owner in NR) Status DNF (bit) Set when FD Flush is not necessary (Future) NodeID (6octets) MC address of message source node (Informational) Reserved(bits), Status Reserved(6bits), Reserved(octets) - Future development Request /State Reserved Status Node ID (6 octets) R D N F Status Reserved (Node ID) Reserved ( octets)

13 Items Under Study G.80 is currently an initial recommendation that will continue to be enhanced. The following topics are under study for future versions of the recommendation: a) Interconnected rings scenarios: shared node, shared links b) blocked at both ends configuration of the ring where both nodes connected to the control the protection mechanism c) Support for Manual Switch administrative decision to close down a link and force a recovery situation are necessary for network maintenance d) Support for Signal Degrade scenarios SD situations need special consideration for any protection mechanism e) Non-revertive mode llows the network to remain in recovery configuration either until a new signal failure or administrative switching f) Displacement Displacement of the role of the to another ring link flexibly in the normal (idle) condition g) In-depth analysis of different optimizations (e.g., FD flushing) h) Etc.

14 Thank You

15 ackup 5

16 G.80 asic Functions Robustness: Source Steering Protection Connectivity: Unicast, MC and C Data Transfers Failure Multicast Flow Unicast Flow 50ms Switch Times 6

17 ERP Functionality Decomposition Local Requests (operator) Local Priority Logic Top Priority Local Request / clear / Clear Local Defect Logic (E) Local Defect Logic (W) ETH_CI_S (E) ETH_CI_S (W) ETH_CI_RPS (E) ETH_CI_RPS (W) Validity Check Request/State + Status Guard Timer Request/State + Status Top Priority Request Priority Logic WTR Running WTR Expire WTR Timer Start/ Stop MI Owner Start/Stop R-PS Request Processing R-PS Request/State + Status Stop Tx R-PS R-PS Message Transmission ETH_CI_RPS (E) ETH_CI_RPS (W) Inputs Local node requests, ETH signals, R-PS messages, WTR Timer. Filters Local Priority Logic, Validity Check, Guard Timer, Priority Logic guarantee that highest priority message arrives for processing. 7 lock/unblock ring ports Flush FD ERP Processing R-PS Request Processing processes filtered message according to State Machine with knowledge of current state and - Owner indication Outputs Node functions (FD Flushing, lock/unblock port), sending (or stopping) R-PS messages

18 State Machine Three node states Initialization when first defining the ring (Idle) the normal state of the ring nodes with blocked and all nodes/ports working (Protecting) when protection switching is in effect, unblocked, other (usually fault) link is blocked. Different input messages assigned priority as appears in the State Machine description. Priority mechanism to allow proper reaction to faults State Machine describes the actions to be taken by the node dependent on current state and input message. Only message with highest priority will pass through the filtering. Different actions include transmission of R- PS message, blocking/unblocking a port, flushing the FD, switching current state of node, and starting/stopping timers. Node State * (Idle) Inputs High Priority request Initialization Local Local Clear R-PS() WTR Expires WTR Running R-PS(NR, R) R-PS(NR) Row # ctions Stop guard timer Stop WTR timer If Owner: lock port Unblock non- port Tx R-PS (NR, R) Else: lock both ports Stop Tx R-PS lock failed port; Unblock non Failed port; Tx R-PS(); Flush FD; No action Unblock non Failed port; Stop Tx R-PS If not DNF Flush FD; No action No action Unblock non- port No action Outputs Next node state 8

19 State Machine FD flushing will clear all of the learned filtering rules within the node. Current Recommendation includes basic optimization only flushing once on fault discovery and once on recovery. DNF status flag is for future definition Node State Inputs High Priority request Local Row # 8 ctions Outputs lock failed port; Unblock non Failed port; Stop WTR; Tx R-PS(); Next node state Local Clear 9 Start guard timer, Tx R-PS(NR), R-PS() 0 Stop WTR, Unblock non Failed port, Stop Tx R-PS (Prote cting) WTR Expires lock port, Unblock non- port Tx R-PS (NR,R) Flush FD, WTR Running No action R-PS(NR, R) If not Owner: Unblock both ports, Stop Tx R-PS, If not DNF Flush FD R-PS(NR) If Owner: Start WTR 9

20 Switching Triggers Detection/Clearing of (Signal Fail): : Loss of CCM, signal passed up from underlying (sub)layer Switching is performed when hold-off timer expires after detection of. Operator s request: Lockout, Forced switch, Manual switch, Clear (for future specification) Remote request: Switching request indicated in received R-PS information from the other node. Protection switching algorithm is based on priorities assigned to all triggers 0

21 Ethernet Rings and the Ethernet nodes connected in a ring topology. Each node connected to two neighbors via links (E & W ports) Deterministic network topology void loops within ring by blocking a designated link () Node responsible to control the blocking state of is owner is unblocked during protection state and may revert to blocked during recovery ETH_FF ETH_C ERP Cntrol Process ERP control ring node ring link ring link ETH_FF ring node ring node C ring protection link () control ETH_C control ERP ERP ERP Control Process ring link connection point (blocked) cont rol ERP ring node D ERP Control Process Owner ETH_FF ETH_C ETH_C ETH_FF ERP Cont rol Process

22 Scenario - Normal to Protection C D E F G Owner Normal failure C D E Flush Flush Flush Flush Flush Flush Flush 50 ms Protection state F G. Normal State. Failure Occurs C. Nodes C and D detect local Signal Failure condition and after respecting hold-off time and lock failed port [row ] D. Node C and D periodically send message, on both ring ports, while condition persists. Each node performs a FD Flush when receiving the message E. Owner receives message and unblocks its end of link [row ] F. Stable State messages on the ring G. Further messages trigger no further action [row0] Message source Client ch block R-PS ch block

23 Scenario recovery Protection state C NR NR D E C D E F G NR failure recovery NR Owner NR NR Confirmation time Normal F G H Flush NR, locked NR, locked Flush Flush Flush Flush Flush NR, locked NR, locked Flush NR, locked NR NR, locked 50 ms. Stable condition Nodes C and D continue to send message every 5 secs.. Recovery of Link failure C. Nodes C and D detect clearing of condition, start guard timer and initiate periodical transmission of NR message on both ring ports. ( Guard timer prevents reception of R-PS messages) [row 9] D. When the Owner receives NR message, it starts the WTR timer [row ] E. When node C and D expire the guard timer, they may accept the new R-PS messages that they receive. [in guard timer] F. t expiration of WTR timer, owner blocks its end of the, sends, NR R message [row ] G. Each node that receives the NR,R message it flushes its FD. When Nodes C and D receive NR R message, they remove block on their blocked ports. [row ] H. Stable normal condition all nodes in Idle state