1. Path Protection in MPLS Networks Ashish Gupta Design and Evaluation of Fault Tolerance Algorithms with Performance Constraints

2. Our Work Fault Tolerance in MPLS Networks Issues QoS Constraints Expeditious Path Restoration Bandwidth Efficiency There is a tradeoff

3. QoS Parameters Important parameters Switch-Over Time End-to-End Delay Reliability Jitter Have to minimize bandwidth usage ADVANCED NETWORKING LAB MPLS PATH PROTECTION

4. Switch-Over Time : Switch-Over Time is the time for which the packets will be dropped in case a failure along the LSP End-to-End Delay : The transmission time of a packet to reach the destination node from the source Reliability : The probabilistic measure of reachability of the destination from the source Jitter : Jitter is the deviation from the ideal timing of receiving a packet at the destination QOS Parameters

5. Path Protection A disjoint backup path is allocated along with the primary path Local Path Protection Global Path Protection Segment Based Approach : A General Approach to Path Protection ADVANCED NETWORKING LAB MPLS PATH PROTECTION

6. Segment Protection Protect each segment separately : Each segment seen as a single unit of failure SSR – Segment Switching router Flexibility in creating segments -> flexibility in Path Protection ( delay and backup paths ) SBPP – Segment Based Path Protection

7. Optimization Problem The structure of backup path(s) and its peering relationship with the primary path affects the QoS Constrains The Design of backup LSPs must address both BW efficiency and expeditious path restoration

8. Explanation of QoS Parameters

9. Switch-Over Time Ensure Switch-Over time RTT( S i, S i+1 ) + T test < delta Where delta is maximum permissible packet loss time

10. End-to-End Delay

11. End-to-End delay Ensure Max (T + ( t2 – t1 ) ) < EED Bound

12. Jitter Ensure Max Jitter from source to destination over all backup paths < Jitter bound

13. Problem Statements

14. Theoretical Model Let G = (R,L) describe the given network where L has the following properties: R = set of routers L = set of links B = Bandwidth of the Links pB = Primary Path bandwidth reserved bB = Backup Path bandwidth reserved D = Delays of the Links P = Reliability

15. Switch-Over Time General Problem Statement Input A Network N, LSP and Switch-over time bound . Output A set of segment switch routers S = Such that S 0 = R 0, S k = R n In case of a fault, the max packet loss time while rerouting is <  RTT ( S i, S i+1 ) + T test <=  No of segments is minimized.

16. Consideration of Backup Paths Input A network N, a LSP and a switch-over time bound  Output A set of segment switch routers S and backup paths { :i=0..k-1} Such that S 0 = R 0, S k = R n In case of a fault, the max packet loss time while rerouting is <  RTT ( S i, S i+1 ) + T test <=  No of segments is minimized.

17. End-to-End Delay General Problem Statement Input A network N, a LSP, switch-over time bound , end-to-end delay bound  Output A set of segment switch routers S and backup paths { :i=0..k} Such that S 0 = R 0, S k = R n In case of a fault, the max packet loss time while rerouting is <  RTT ( S i, S i+1 ) + T test <=  No of segments is minimized. Backup path constraints

18. Jitter General Problem Statement Input A network N, a LSP, switch-over time bound , jitter bound J Output A set of segment switch routers S and backup paths { :i=0..k} Such that S 0 = R 0, S k = R n In case of a fault, the max packet loss time while rerouting is <  RTT ( S i, S i+1 ) + T test <=  No of segments is minimized. Backup path constraints Jitter J

19. General Problem Statement Input A Network G and Packet Loss time bound delta and jitter bound delta j. an ingress Node a and an egress node b between which a connection of bandwidth y has to be routed. Output A primary path between a and b, a set of segment switch routers S and set of backup paths BP. Such that S 0 = a In case of a fault, maximum jitter bound is delta j Max ( t2 – t1 ) < delta j RTT ( S i, S i+1 ) + T test <= delta Bandwidth resources are conserved No of segments is minimized or |S| is minimum( Transformation )

20. Algorithm d1d1 d2d2 d3d3 d 1 + d 2 + d 3 d3d3 0 d 2 + d 3

21. Reliability General Problem Statement Input A network N, a LSP, switch-over time bound , minimum reliability requirement r Output A set of segment switch routers S and backup paths { :i=0..k} Such that S 0 = R 0, S k = R n In case of a fault, the max packet loss time while rerouting is <  RTT ( S i, S i+1 ) + T test <=  No of segments is minimized. Backup path constraints Minimum reliability is r

22. RELIABILITY - 1 How Backup Path Improves Reliability Link Reliability : p e n links each in the primary and backup paths. Reliability from A to B without a backup path = p Reliability from A to B with backup path = 2 p – p 2

23. RELIABILITY - 2

24. RELIABILITY - 3 How Backup Path Improves Reliability Link Reliability : p e n links each in the primary and backup paths. Reliability from A to B without a backup path = p n Reliability from A to B with backup path = 2 p n – p 2*n A B

25. RELIABILITY - 4 Segment Heads Backup Paths Total number of links in primary path = n Size of Backup Path = Size of Segment Size of Segments = k Assume no sharing of backup paths

26. RELIABILITY - 5 Reliability of a link : p Reliability of a segment = 2p k – p 2k Number of Segments = n/k Reliability of the path = (2p k – p 2k ) n/k

27. RELIABILITY – 6

28. Algorithm How to calculate reliability Given segment heads, find the most reliable backup paths Find segment heads

29. How to Calculate Reliability? NP-Complete problem, even when failure probability is same for all links. For a graph G with edge reliability p e for edge e, where O is the set of operational states. Therefore we will have to estimate reliability of a path by using upper and lower bounds.

30. Graph Transformations Node to Link Reliability A pnpn A1A1 A2A2 pnpn Merging Serial Parallel pepe pfpf P e *p f pepe pfpf p e + p f - p e *p f

31. Approximating Reliability Consider a path from link A to B Total number of links in primary and backup paths = n Reliability of a link : p Probability ( failure of k links ) n c k * p n-k * (1-p) k

32. Probability of k links failing Probability that 0 or 1 or 2 links failed = 0.9861819

33. Approximating Reliability

34. Number of States with 0 link failure : n c 0 Probability of occurrence of this state : p n Probability that a path exist : 1 Number of States with 1 link failure : n c 1 Probability of occurrence of this state : p n-1 (1-p) Probability that a path exist : 1 Number of States with 2 link failure : n c 2 Probability of occurrence of this state : p n-2 (1-p) 2 Probability that a path exist : From Simulation(say q)

35. Approximating Reliability Lower Bound n c 0 * p n * 1.0 + n c 1 * p n-1 (1-p) * 1.0 + n c 2 * p n-2 (1-p) 2 * q Upper Bound 1 - n c 2 * p n-2 (1-p) 2 * (1-q)

36. Lower & Upper Bounds

37. Reliability

38. Finding Reliable Backup Paths R1R1 R5R5 R6R6 R7R7 R8R8 R9R9 R 10 R 11 R 12 R2R2 R3R3 R4R4 r 9 12 r 10 12 r 11 12 1 Given the segment heads, we can find backup paths that maximizes reliability of the network.

39. Finding Segment Heads Approach #1 Consider all possible segmentations. Approach #2 Find the best possible segmentation without considering reliability while segmenting. Divide segments to improve reliability till reliability becomes greater than required.

40. Algorithm Which segment to divide first? Divide segment with maximum reliability first Divide longest segment first Random

41. Future Work Algorithm for protection meeting reliability criteria Optimization issues – Bandwidth, capacity Implementation of these algorithms in emulator and experimental setup