wilkinsontalk

wilkinsontalk - Scalable Parallel Primitives for Massive...

Info iconThis preview shows pages 1–12. Sign up to view the full content.

View Full Document Right Arrow Icon
1 Scalable Parallel Primitives for Massive Graph Computation !"#$% Buluç University of California, Santa Barbara
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Sources of Massive Graphs (WWW snapshot, courtesy Y. Hyun) (Yeast protein interaction network, courtesy H. Jeong) Graphs naturally arise from the internet and social interactions Many scientific (biological, chemical, cosmological, ecological, etc) datasets are modeled as graphs .
Background image of page 2
Examples: - Centrality - Shortest paths - Network flows - Strongly Connected Components Examples: - Loop and multi edge removal - Triangle/Rectangle enumeration 3 Types of Graph Computations Fuzzy intersection Examples: Clustering, Algebraic Multigrid Tool: Graph Traversal Tool: Map/Reduce 1 3 4 2 5 7 6 Tightly coupled Filtering based map map map reduce reduce
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Tightly Coupled Computations Many graph mining algorithms are computationally intensive. (e.g. graph clustering, centrality) Some computations are inherently latency-bound. (e.g. finding shortest paths) Interesting graphs are sparse, typically |edges| = O(|vertices|) 4 Huge Graphs Expensive Kernels + ! High Performance and Massive Parallelism Sparse Graphs/Data Sparse Data Structures (Matrices) ! Tightly Coupled Computations on Sparse Graphs
Background image of page 4
5 Software for Graph Computation '((()" )+,% -.%-/01,.% +2345 spending ten years of my life on the TeX project is that software is hard. It's harder than anything 4/14 6784 4845 9+# 3. #.:
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
6 Software for Graph Computation '((()" )+,% -.%-/01,.% +2345 spending ten years of my life on the TeX project is that software is hard. It's harder than anything 4/14 6784 4845 9+# 3. #.: Dealing with software is hard !
Background image of page 6
7 Software for Graph Computation '((()" )+,% -.%-/01,.% +2345 spending ten years of my life on the TeX project is that software is hard. It's harder than anything 4/14 6784 4845 9+# 3. #.: Dealing with software is hard ! High performance computing (HPC) software is harder !
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
8 Software for Graph Computation '((()" )+,% -.%-/01,.% +2345 spending ten years of my life on the TeX project is that software is hard. It's harder than anything 4/14 6784 4845 9+# 3. #.: Dealing with software is hard ! High performance computing (HPC) software is harder ! Deal with parallel HPC software?
Background image of page 8
Outline The Case for Primitives The Case for Sparse Matrices Parallel Sparse Matrix-Matrix Multiplication Software Design of the Combinatorial BLAS An Application in Social Network Analysis Other Work Future Directions 9
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
10 Input: ;,54-34# <5+=9 >,39 '-.131: .% 4#<41 Find least-cost paths between all reachable vertex pairs Classical algorithm: Floyd-Warshall Case study of implementation on multicore architecture: ? graphics processing unit (GPU) All-Pairs Shortest Paths for k =1:n // the induction sequence for i = 1:n for j = 1:n if( w(i ! k)+w(k !" ) < w(i !" ) ) w(i !" ):= w(i ! k) + w(k !" ) 1 5 2 3 4 1 5 2 3 4 k = 1 case
Background image of page 10
11 GPU characteristics Powerful: two Nvidia 8800s > 1 TFLOPS Inexpensive: $500 each ! 3/1 "/4'*%$'"+/ 4'*1-. 0*"514 6 -*"'7.1'"$ %/"'4
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 12
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 40

wilkinsontalk - Scalable Parallel Primitives for Massive...

This preview shows document pages 1 - 12. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online