1. Pipelined Broadcast on Ethernet Switched Clusters Pitch Patarasuk, Ahmad Faraj, Xin Yuan Department of Computer Science Florida State University Tallahassee, FL 32306

2. Broadcast communication(MPI_Bcast) n0n0 n1n1 n2n2 n3n3 n0n0 n1n1 n2n2 n3n3 Before After ABCD ABCDABCDABCDABCD Let T(msize) = time to send a message of size msize Broadcast(msize) >= T(msize)

3. Ethernet Switched Cluster switch

4. Problem statement: How to efficiently realize the broadcast operation with large message sizes on Ethernet switched clusters. Using pipelined broadcast can achieve near optimal results (T(msize) time for broadcasting a message of size msize). Finding contention free broadcast tree Finding a good segment size

5. Traditional Broadcast algorithms 01234567 Linear tree 1234567 Flat tree 0 Time = (P-1) x T(msize)

6. 0 12 3456 7 Binary tree 0 123 4567 k-ary tree Time = 2x(log 2 (P+1)-1)xT(msize)

7. 0 42 65 1 3 7 Binomial tree Time = log 2 P x T(msize)

8. Scatter/Allgather n0n0 n1n1 n2n2 n3n3 Before ABCD ABCD Scatter Allgather ABCDABCDABCDABCD Time = 2 x T(msize)

9. Time Complexity for large messages Linear tree(P-1) x T(msize) Flat tree(P-1) x T(msize) Binary tree2x(log 2 (P+1)-1)xT(msize) Approx. 2xlog 2 P x T(msize) Binomial treelog 2 P x T(msize) Scatter/allgather2xT(msize)

10. Pipelined Broadcast Algorithm Linear pipeline 0123

11. Performance of pipelined broadcast: Assume no network contention a message of size msize be broken into X messages of msize/X. H: tree hight, D: the number of children Size of pipelined stage: D * T(msize/X) Total time T: (X + H –1) * (D * T(msize /X)) linear tree: H = P, D = 1, T = T(msize) Binary tree: H = log(P), D= 2, T = 2T(msize) K-ary tree: H = log_k(P), D = k, in general not as efficient as binary tree.

12. Time Complexity for large messages Pipelined (linear)T(msize) Pipelined (binary)2 x T(msize) k-ary pipelinek x T(msize) Binomial treelog 2 P x T(msize) Scatter/allgather2xT(msize)

13. Pipelined broadcast How to find a contention-free broadcast tree? How to select the best segment size?

14. Example of network contention 0 12 3456 7 Binary tree switch n 0,n 1,n 2,n 3 n 4,n 5,n 6,n 7 There is a link contention cause by communication (1  4), (2  5), (2  6), and (3  7)

15. Linear tree switch n 0,n 1,n 4,n 5 n 2,n 3,n 6,n 7 The linear tree 0  1  2  3  …  7 will have a contention caused by (1  2) and (5  6)

16. Algorithm for constructing contention free linear tree Step 1: Traverse through all switches using depth-first-search (DFS) algorithm, name the switch by the order of their arrival in DFS tree Step 2: The linear tree consists of all machines in switch S 0, follows by all machines in S 1, then S 2,and so on

17. Example of contention free linear tree Switch S0 Switch S1 n 0,n 1,n 4,n 5 n 2,n 3,n 6,n 7 Switch S3 Switch S2 n 12,n 13,n 14,n 15 n 8,n 9,n 10,n 11 Linear tree: n0  n1  n4  n5  2  3  6  7  8  9  …  15

18. Algorithm for constructing contention free binary tree Start with a contention free linear tree Recursively divide the tree into 2 sub-trees Make sure that the cannot be a contention The sub-trees are chosen such that the height of the whole tree will be minimal 0123456789101112131415

19. Binary tree height Performance of binary pipeline broadcast depends on the height of a binary tree Even though contention free binary tree may not be a complete binary tree, its height is not that much more than a complete binary tree

20. Average tree heights for 20 randomly generated topologies

21. Evaluation Contention free pipelined algorithms: Routine generators from topology information The generated routines are based on MPICH p2p primitives. Linear tree Binary tree 3-nary tree Targets for comparison: MPICH: Binomial tree, Scatter/allgather LAM: Flat-tree, Binomial Topology unaware pipelined linear and binary algorithms

22. Evaluation

23. Performance of different pipelined trees (topology 1)

24. Comparing pipelined broadcast with other schemes

25. Topology unaware and contention-free pipelined broadcast

26. Segment size for pipelined broadcast

27. Conclusions Pipelined broadcast is faster than the current broadcast algorithm for medium and large messages Linear pipeline has a completion time roughly equal to T(msize) binary pipeline broadcast is best for medium messages Contention free broadcast tree is necessary for pipelined algorithms A good segment size for pipelined broadcast is not difficult to find.

28. Questions?