1 / 47

LSRP: Local Stabilization in Shortest Path Routing

LSRP: Local Stabilization in Shortest Path Routing. By Anish Arora Hongwei Zhang Presented By: Ankit Malhotra ankit@virginia.edu. Local Self Stabilization : Need. Network. Local Self Stabilization : Need. Fault. Fault Propagation Wave. Network. Local Self Stabilization : Need.

lev
Télécharger la présentation

LSRP: Local Stabilization in Shortest Path Routing

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. LSRP: Local Stabilization in Shortest Path Routing By Anish Arora Hongwei Zhang Presented By: Ankit Malhotra ankit@virginia.edu

  2. Local Self Stabilization : Need Network

  3. Local Self Stabilization : Need Fault Fault Propagation Wave Network

  4. Local Self Stabilization : Need Corrected Node Correction Wave Network Fault Propagation Wave

  5. Local Self Stabilization : Need Network Correction Wave Fault Propagation Wave

  6. Local Self Stabilization : Need Network Fault Propagation Wave Correction Wave

  7. Local Self Stabilization : Need Hence Unbounded Fault Propagation is not acceptable. Because it will affect large part (if not whole) system before it is rectified. Network Correction Wave

  8. Overview • The Basics • Local Stabilization – Definitions and Properties • LSRP Protocol • Conclusion

  9. The Basics • The protocol is written using the guarded command notation <name> : <guard> ------> <statement> e.g. S1 : (i=r & d.i=0) | (ghost.i & SP.i) -----> p.i=i • At system state q, • System topology : G.q(V.q,E.q) • State of node i : q(i)

  10. Local Stabilization • F-Local Stabilization • Faults be contained locally around where they occurred. • Time taken for the system to stabilize is a function of the size of the perturbed region. Locally Contained Fault Regions Correction Definite Time which is proportional to size of perturbed region

  11. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) Implies, the set of nodes dependant on V’ and E’ at a legitimate system state q.

  12. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) 0 1 1 2 2 For this node, DS is 3 3 4 4 4 5 5 5 6

  13. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) Equals to the minimum number of nodes whose states either have been corrupted or whose variables have to be changed for stabilization

  14. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) State Corruption 0 0 1 1 1 1 2 2 2 2 Node Failure 2 3 3 4 4 4 4 4 4 5 5 5 5 5 5 6 6 Perturbation Size = 1 Perturbation Size =3

  15. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) The set of node sets that can be potentially corrupted because of a node failure or state corruption in their parent tree.

  16. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) State Corruption 0 0 1 1 1 1 2 2 2 2 Node Failure 2 3 3 4 4 4 4 4 4 5 5 5 5 5 5 6 6 PP(q) PP(q)

  17. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) The set of nodes that are perturbed in state q. It is a subset of the Potentially Perturbed Nodes, PP(q).

  18. Local Stabilization (Contd.) • Definitions • Dependent Set, DSq(V’,E’) • Perturbation Size, P(q) • Potentially Perturbed Nodes, PP(q) • Perturbed Node Set, PN(q) State Corruption 0 0 1 1 1 1 2 2 2 2 Node Failure 2 3 3 4 4 4 4 4 4 5 5 5 5 5 5 6 6 PN(q) PN(q)

  19. Local Stabilization (contd) • Properties of such systems • maximum distance the faults can propagate outwards is O(F(P(q))) • local stabilization of one region is independent and concurrent of the others. • availability of G is high • guarantee local fault containment with repeated faults.

  20. Problem Statement • Design a protocol that given a system G, constructs and maintains a spanning tree T of G such that: • Destination ‘r’ is the root of T • T is the shortest path tree • G is F-local stabilizing Node ‘r’

  21. Existing Solutions • Faults cannot be contained

  22. Existing Solutions • Faults cannot be contained F-Local Stabilization is not guaranteed

  23. Existing Solutions • Faults cannot be contained  F-Local Stabilization is not guaranteed  routing instability

  24. The ProblemandSuggested Solution • Node of Fault Propagation to initiate a “Containment” action that moves faster than the stabilization (“Fault Propagation”) action. • “Corrective” action always lags behind “Fault propagation” action Correction Wave Fault Propagation Wave Containment Wave

  25. The ProblemandSuggested Solution • Node of Fault Propagation to initiate a “Containment” action that moves faster than the stabilization (“Fault Propagation”) action. • “Corrective” action always lags behind “Fault propagation” action Correction Wave Fault Propagation Wave Containment Wave

  26. The ProblemandSuggested Solution • Node of Fault Propagation to initiate a “Containment” action that moves faster than the stabilization (“Fault Propagation”) action. • “Corrective” action always lags behind “Fault propagation” action Correction Wave Fault Propagation Wave Containment Wave

  27. The ProblemandSuggested Solution • Node of Fault Propagation to initiate a “Containment” action that moves faster than the stabilization (“Fault Propagation”) action. • “Corrective” action always lags behind “Fault propagation” action Correction Wave Fault Propagation Wave Containment Wave

  28. The ProblemandSuggested Solution • Node of Fault Propagation to initiate a “Containment” action that moves faster than the stabilization (“Fault Propagation”) action. • “Corrective” action always lags behind “Fault propagation” action Correction Wave Fault Propagation Wave Containment Wave

  29. One Problem • It could happen that the containment wave is mistakenly initiated by state corruption. To prevent that the super containment wave is introduced. • It is essential, that the Super Containment Wave must self stabilize itself locally upon perturbations. • This is achieved by ensuring that the super containment wave only uses the variables defined for the stabilization wave and containment wave.

  30. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC r: The ID of the destination node ds: Inverse of Propagation speed of stabilization wave dc: Inverse of Propagation speed of stabilization wave dsc: Inverse of Propagation speed of stabilization wave L & U: any constants such that 0<=L<=U. Stabilization Containment Super Containment

  31. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC d.i: shortest distance from i to r. p.i: next hop on the shortest path from i to r. ghost.i: boolean variable for knowing if i is involved in containment wave or not. N.j: The neighbohood set of j k: dummy variable Stabilization Containment Super Containment

  32. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC S1: If node ‘i’ is a minimal node but p.i != i, set p.i = i Stabilization 1 0 1 1 0 1 2 2 Containment Super Containment 0 0 1 1 1 1 2 2

  33. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC S2: If node i should propagate a stabilization wave from j, that is not being involved in any containment wave, then i sets j as its parent and updates d.i, ghost.i to d.j+1, false respectively Stabilization Containment Super Containment

  34. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC 0 Here node j should initiate a stabilization wave from j, So it updates d.i and ghost.i 1 1 2 2 j Stabilization 1 3 i 4 4 4 5 5 5 Containment 0 6 1 1 2 2 Super Containment j 1 3 i 4 4 2 5 5 5 6

  35. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC C1: If ghost.i = false, but it is a source of fault propagation, or it should propagate a containment wave from its parent, then i sets ghost.i=true. Also if i is the source of fault propagation, it sets p.i=i Stabilization Containment Super Containment

  36. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC 0 Here node j realizes that d.j is shortest amongst its neighbors, so it is the source of fault propagation and hence it initiates a containment wave. 1 1 2 2 j Stabilization 1 3 ghost.j=false i 4 4 2 5 5 5 Containment 6 0 1 1 2 Super Containment 2 j 1 3 i ghost.j=true 4 4 2 5 5 5 6

  37. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC • C2: If node i is involved in a containment wave, but i has no child k that is perturbed due to state corruption at i, then i sets ghost.i=false and • If i is r, then d.i=0 and p.i=I • If there exists a parent substitute of i, then d.i=d.j+1 and p.i=j, else • d.i=∞ and p.i=i Stabilization Containment Super Containment

  38. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC 0 Here j’ serves as a parent substitute of j. 1 1 j’ 2 2 j Stabilization 1 3 i ghost.j=true 4 4 2 5 5 5 Containment 6 0 1 1 j’ 2 Super Containment 2 j 3 3 i ghost.j=false 4 4 2 5 5 5 6

  39. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC • SC: If node i is involved in a containment wave and, • It is the destination node and d.i = 0, or • Neither it is a source of fault propagation nor is its parent involved in any containment wave. • then, i sets ghost.i=false. Stabilization Containment Super Containment

  40. Protocol Design • Constants : r, ds, dc, dsc, L, and U • Variables : d.i, p.i, ghost.i, N.i, k • Actions • S1 • S2 • C1 • C2 • SC 0 This value of node j will make i become a source of fault propagation and initiate a containment wave. But after initiating the containment wave it will realize that it should’nt. Hence it would initiate a super containment wave which will stop the mistakenly initiated containment wave. 1 1 2 Stabilization 2 j 4 3 i 4 4 4 Containment 5 5 5 Super Containment 6

  41. SPEED? Is it F-Local Stabilizing

  42. SPEED? Is it F-Local Stabilizing • Yes, • Since SPEED only takes the routing decisions till the best neighbor and not destination, any faults at a node, x will be contained in the neighborhood of x only. Hence it is F-local, • Also since SPEED does have a inherent stabilization function (Beacon Exchange) it will self stabilize. Hence SPEED is F-Local Stabilizing. • In fact we can generalize, any protocol that makes its decision based only its immediate neighbors without keeping the destination in context is F-Local Stabilizing if they have a local stabilizing function.

  43. Conclusion and Discussion • What happens if the perturbed regions overlap?

  44. Conclusion and Discussion • What happens if the perturbed regions overlap? • What happens if there happens a state corruption? after the super containment wave is sent out, which changes the value of ghost.i back to true?

  45. Conclusion and Discussion • What happens if the perturbed regions overlap? • What happens if there happens a state corruption? after the super containment wave is sent out, which changes the value of ghost.i back to true? • How can u prove that the super containment wave cannot go out because of some state corruption?

  46. Conclusion and Discussion • What happens if the perturbed regions overlap? • What happens if there happens a state corruption? after the super containment wave is sent out, which changes the value of ghost.i back to true? • How can u prove that the super containment wave cannot go out because of some state corruption? • The paper puts forth that the algorithm described for self stabilizing is applicable in general. How?

  47. Thank You

More Related