1. Routing Without Flow Control Hot-Potato Routing Simulation
Project:
Routing Without Flow Control Hot-Potato Routing Simulation
by Larry Bush

2. Motivation Paper Summary

3. Overview Concepts How Algorithm Works Demonstration Results
More than one processor, each with a its own cursor, can be traversing and manipulating the linked list.
Present algorithms and data structures that directly implement a non-blocking singly-linked list.
Allows multiple processes to traverse, insert and delete.
Using only commonly available Compare&Swap.

4. Algorithm Concepts
Flow Control
Hot Potato Routing
Greedy
Dynamic

5. Topology
N*N Torus Network

8. The Algorithm Good-Link / Bad-Link Home-Run Path 4 priority states:
Sleeping … Lowest Priority
Active
Excited
Running … Highest Priority

9. State Change Rules
Sleeping
Active
Excited
Running

10. Active Packet Demo

30. Running Packet Demo

51. Conflicts
Both Packets
Conflicts

78. Demo
Both Packets
Plus Other Sleeping Packets

105. Guarantees
O(N) Delivery Time
O(N) Injection Time

110. Conclusion Algorithm behaved as expected. ROSS Speed-up is Good.
There are currently universal (for any ADT) wait free methods but they have too much overhead to be efficient. Therefore, this paper shows a direct implementation that is more efficient.
Valois’ uses a single word version of Compare&Swap which is commonly available on most systems.

111. Take Aways Difficult to analyze theoretically.
Simulation is a good tool for this type of problem.
ROSS is a good tool for this type of simulation.

112. End
There are currently universal (for any ADT) wait free methods but they have too much overhead to be efficient. Therefore, this paper shows a direct implementation that is more efficient.
Valois’ uses a single word version of Compare&Swap which is commonly available on most systems.

113. Questions/Facts

114. Theoretical Model vs. Practical Model
Does not Absorb Sleeping Packets
Practical
Absorbs Sleeping Packets

115. Outline (notes) Introduction to the problem Project scope. Motivation.
Utilized all free links.
Paper Overview.
Define concepts in model.
Define model.
Destinations chosen randomly
Demonstrate model.
Results
Model results.
Explain Graphs
Delivery Time (Theoretical Model / Practical Model)
Injection Time (Theoretical Model / Practical Model)
Summarize Results
ROSS performance
Speed-up (Parallel vs. Sequential)
Event Rate
Execution Time
Parameter Effects
Remote Events Sent / Total Events Processed
Roll-backs / Net Events

116. Graphs / Results Results Model results. ROSS performance
Explain Graphs
Delivery Time
Injection Time
ROSS performance
Speed-up (Parallel vs. Sequential)
Event Rate

117. Project Scope. - Sim & Analyzed. - Algo
Project Scope - Sim & Analyzed - Algo. In Costas’ Paper Paper Overview - New version of HPR. - Proved Theoretically. - Dynamic vs. Static Motivation - Utilized all free links. - Difficult to analyze theoretically. Simulation is a better tool for this type of problem.

118. Network Concepts Synchronous Buffer-less Model of Optical
Switching Network