150 likes | 351 Views
Parallel Skyline Computation on Multicore Architectures. ICDE`09. Outline. Introduction Preliminary Parallel BBS(branch-and-bound algorithm) Parallel Skyline Algorithm ( Pskyline ) Experiments Conclusion. Introduction. (cont.). The advantage of the skyline Skyline algorithm Sequential
E N D
Parallel Skyline Computation on Multicore Architectures ICDE`09
Outline • Introduction • Preliminary • Parallel BBS(branch-and-bound algorithm) • Parallel Skyline Algorithm (Pskyline) • Experiments • Conclusion
(cont.) • The advantage of the skyline • Skyline algorithm • Sequential • No index structure • BNL、SFS 、LESS • Index structure • NN 、BBS 、ZSHARE • Parallel • Many in Distributed environment, but no algorithm for multi-core environment
Prrliminary • A. Skyline computation • For a d-dimensional dataset D. • skyline set : • Incomparable : • Transitivity : • Incomparability : • Distributivity :
(cont.) • B.Skeletal parallel programming • A programming model • Need two component : • Data structure • List : • Parallel skeletons • not • For developing parallel skyline algorithm • Parallel map and Parallel reduce
(cont.) • Pmap (parallel map) A B C D E F G H A’ B’ C’ D’ E’ F’ G’ H’
(cont.) • Preduce (parallel reduce) A B C D E F G H A’ B’ C’ D’ …… R
Parallel BBS • A. branch-and-bound algorithm • Use R-tree as its index structure
(cont.) • B.Parallelizing BBS
Parallel skyline algorithm (Pskyline) • Overall design • Pskyline D = S(D) • PskylineD = sreducepmerge (pmapsskyline L) • L = [D1,…,Db] D = D1++…++Db • Sreduce
(cont.) • Pmerge • Sskyline
Conclusion • Muliti-core architecture for database operations.