Transcript Dijkstra-Scholten and Shavit-Francez termination algorithms

Dijkstra-Scholten and Shavit-Francez
termination algorithms
Index :
•
•
•
•
•
Introduction
Experimental Setup
Result Analysis
Conclusion
Future Work
Introduction
• Dijkstra-Scholten algorithm detects the termination of
a centralized basic computation.
• Shavit-Francez generalized Dijkstra – Scholten to
decentralized basic computations (multiple initiators)
• In this algorithm, each initiator maintains a
computation tree similar to Dijkstra-Scholten and
when the computation tree collapses it may
participate in the computation tree of other initiators.
• All processes participate in Wave to detect the
termination of computation.
Experimental Setup
• Echo algorithm runs on the top of basic algorithm.
• Each node participate in computation at least once.
• Algorithm terminates when Echo algorithm decides.
• Total number of messages (basic + control)
exchanged in the algorithm are counted.
Result Analysis :Only one process
participate in computation as initiator
3000
2500
2000
Fully connected
1500
Ring
Star
1000
500
0
5
10
15
20
25
30
35
40
45
50
10000
Result Analysis :50% of process
participate in computation as initiators
5000
2500
2000
1500
Fully connected
Ring
Star
1000
500
0
5
10
15
20
25
30
35
40
45
50
Result Analysis :All process participate
in computation as initiators
25000
20000
15000
Fully connected
Ring
Star
10000
5000
0
5
10
15
20
25
30
35
40
45
50
10000
Conclusion
• Observed behavior of Shavit-Francez termination
algorithm for different topologies.
Future Work
• Perform the same analysis for various wave
algorithms.
• To observe performance of algorithm when random
nodes participate in computation.