Load balancing in large-scale epidemiological simulations

Authors:
Tariq Kamal;Keith R. Bisset;Ali R. Butt;Youngyun Chungbaek;Madhav Marathe
Affiliations:
Virginia Tech, Blacksburg, VA, USA;Virginia Tech, Blacksburg, VA, USA;Virginia Tech, Blacksburg, VA, USA;Virginia Tech, Blacksburg, VA, USA;Virginia Tech, Blacksburg, VA, USA
Venue:
Proceedings of the 22nd international symposium on High-performance parallel and distributed computing
Year:
2013

Citing 3
Cited 0

A Fast and High Quality Multilevel Scheme for Partitioning Irregular Graphs

SIAM Journal on Scientific Computing
EpiSimdemics: an efficient algorithm for simulating the spread of infectious disease over large realistic social networks

Proceedings of the 2008 ACM/IEEE conference on Supercomputing
Simulating the Spread of Infectious Disease over Large Realistic Social Networks Using Charm++

IPDPSW '12 Proceedings of the 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops & PhD Forum

Quantified Score

Hi-index	0.00

Visualization

Abstract

Despite the recent advancements in graph partitioning techniques and algorithms, achieving static load balancing in agent-based epidemiological applications is challenging. Input to these simulations is a large agent-location bipartite graph that is highly complex and non-uniform. In this paper, we compare several static load distribution schemes, including our custom strategies, for partitioning a class of bipartite graphs. Computations over such graphs happen between classes of nodes in phases. Our performance evaluations on a 768 core system show that our lower-overhead custom load balancing strategy achieves a 2-fold increase in strong scaling performance compared to the default Round Robin data distribution.