LSA Flooding Optimization Algorithms and Their Simulation Study (draft-choudhury-manral-flooding-simulation-00.txt)

1
LSA Flooding Optimization Algorithms and Their
Simulation Study (draft-choudhury-manral-flooding
-simulation-00.txt)
Gagan Choudhury ATT gchoudhury_at_att.com
Vishwas Manral NetPlane Systems VishwasM_at_netplane
.com
2
The Basic Issue

• Flooding Over All Interfaces is Highly Reliable
  But in Large Networks it May Cause Sustained CPU
  Congestion (Often Memory Congestion as well)
  During LSA Storms Triggered By
• Links/Nodes Failures
• Synchronization of Refreshes
• Software Bugs or Procedural Errors
• Congestion Reinforced by Positive Feedback Loop
  due to
• LSA Retransmissions, possible packet droppings,
  possible link failures due to missed Hellos and
  eventual recoveries More LSAs
  
• On Rare Occasions the Congestion Spreads to Many
  Nodes and Cause Significant Failures (Observed in
  Operational Networks)
• We Show Simulation Study on How
  Stability/Scalability of Networks May Be Improved
  with Restrictive Flooding Algorithms and Propose
  that a Subset of These Schemes be Pursued Further

3
Flooding Algorithms

• Algorithm 1 Flood over All Interfaces (Existing
  Algorithm)
• Algorithm 2 Full Flooding But Flood over Only
  one of Many Parallel Links Between Neighbors
  (Zinin/Shand ID, Moy ID, Used in PNNI)
• Algorithm 3 Algorithm 2 Full Flooding only at
  Multipoint Relays Chosen by Each Node
  Independently
• Algorithm 4 Algorithm 2 Flooding only Along a
  Minimum Spanning Tree (If a Link Along the Tree
  Fails the MST Needs to be Re-computed) Not
  Robust Under Failures
• Algorithm 5 Algorithm 2 for LSAs Carrying
  Intra-Area Topology (Router, Network), Algorithm
  4 for Other LSAs (ASE, TE, Summary)
• Modified Algorithm 5 Flooding Links Survivable
  Under Single Link and Single Node Failures
  (Results Not Reported)

4
Alternate Simulation Scenarios

• Network Scenarios
• Network 1 100 Nodes, 1200 Links, Max Neighbors
  30, Max Node Adjacency 50
• Network 2 50 Nodes, 600 Links, Max Neighbors 25,
  Max Node Adjacency 48
• LSA Scenarios
• 1 Router LSA per Node, 1 TE LSA per Link
• 1 Router LSA per Node, 10 ASE LSAs per Every
  Other Node
• LSU Processing Time 1 ms, 0.5 ms, 2 ms

5
Five Simulation Cases

• Case 1 Network 1, Link LSAs, Proc. Time 1 ms
• Case 2 Network 1, ASE LSAs, Proc. Time 1 ms
• Case 3 Network 1, Link LSAs, Proc. Time 0.5
  ms
• Case 4 Network 1, Link LSAs, Proc. Time 2 ms
• Case 5 Network 2, Link LSAs, Proc. Time 1 ms

6
Number of Non-Converged LSAs Vs. LSA Storm -
Case 1, Algorithm 1 - LSA Storm Starts
Between 20 and 30 Seconds
7
LSA Storm Threshold for Sustained CPU Congestion
8
Observations on Flooding Algorithms

• Flooding Over one of Many Parallel Links (Alg. 2)
  May Significantly Improve Scalability over
  Current Algorithm (Alg. 1)
• Zinin/Shand ID/ Moy ID Should be Pursued Further
• Further Restriction With MPR (Alg 3) Has Moderate
  Improvement
• Neighbors of High Adjacency Node Tend to Declare
  it as MPR
• Flooding Only Over Minimum Spanning Tree (Alg 4)
  Greatly Improves Scalability But Not Robust Under
  Failure
• Full Flooding for LSAs Carrying Intra-Area
  Topology and MST Flooding for Others (Alg 5) May
  Be Almost As Scalable as Alg 4 But Also More
  Robust
• Modification to Alg 5 with Disjoint MSTs Other
  Flooding Links to Ensure Robustness Under Single
  Link and Single Node Failures Have Been
  Considered (Not Reported)
• Quite Robust and Significantly More Scalable
  Compared to Alg 2
• Alg 2, Alg 5 and Modified Alg 5 Should Be Pursued
  Further