P A R A L L E L A L G O R I T H M S F O R M I N I N G L A R G E - - PowerPoint PPT Presentation

P D S W- D I S C S 2 0 1 8 : 3 R D J O I N T I N T E R N AT I O N A L W O R K S H O P O N PA R A L L E L D ATA S T O R A G E & D ATA I N T E N S I V E S C A L A B L E C O M P U T I N G S Y S T E M S P A R A L L E L A L G O R I T H M S F O R M I N I N G L A R G E - S C A L E T I M E - V A R Y I N G ( D Y N A M I C ) G R A P H S Big Data and S H A I K H A R I F U Z Z A M A N Scalable Computing N AW S A F R I N S AT TA R Research Lab M D A B D U L M OTA L E B F AY S A L New Orleans, LA 70148 USA

TEMPORAL GRAPH • powerful representation of various social, biological and technological dynamic systems – Social interactions and human activities, – appearance and disappearance of links in the Web, – patterns of interactions among genes – patterns of interactions in functional brain networks • Complex system Big Data and Scalable Computing Research Lab 2

APPLICATIONS • Diffusion and propagation in complex and social networks – spread of viruses through a community • Understanding communication networks – false news propagation • Improving transportation systems – route-planning algorithms depending on the traffic with varied time • Neuron (brain) network analysis – Locating key neurons in cortical networks Big Data and Scalable Computing Research Lab 3

MOTIVATION & CHALLENGES • defining and computing various temporal network metrics –classic studies’ analysis of the topological properties of static graphs • emergence of network big data –massive networks often do not fit in the main memory of a single machine –prohibitively large runtime for existing sequential methods Big Data and Scalable Computing Research Lab 4

OUR APPROACH • Designing scalable algorithms • Metrics –Computing path –Centrality –Communities Big Data and Scalable Computing Research Lab 5

OUR SCALABLE ALGORITHMS FOR STATIC GRAPH DISTRIBUTED-MEMORY DPLAL (DISTRIBUTED PARALLEL PARALLEL ALGORITHMS FOR LOUVAIN ALGORITHM WITH COUNTING AND LISTING LOAD-BALANCING) TO DETECT TRIANGLES IN BIG GRAPHS COMMUNITIES Big Data and Scalable Computing Research Lab 6

OUR MULTI-THREADED ALGORITHM FOR DYNAMIC GRAPH SPEEDUP VS NO OF THREADS Speedup factors of youtube string biogrid our multithreaded 3.5 3 shortest path 2.5 SPEEDUP algorithm 2 1.5 1 0.5 0 1 3 5 7 NO OF THREADS Big Data and Scalable Computing Research Lab 7

FUTURE WORKS Parallelize existing sequential temporal networks mining and computation of network metrics –Efficient load balancing –Communication schemes –Data reduction (e.g., graph sparsification and approximation) –Efficient formalization of temporal metrics Big Data and Scalable Computing Research Lab 8

THANK YOU Big Data and Scalable Computing Research Lab 9