The Potential of Diffusive Load Balancing at Large Scale
Research output: Contribution to conferences › Paper › Contributed › peer-review
Contributors
Abstract
Dynamic load balancing with diffusive methods is known to provide minimal load transfer and requires communication between neighbor nodes only. These are very attractive properties for highly parallel systems. We compare diffusive methods with state-of-the-art geometrical and graph-based partitioning methods on thousands of nodes. When load balancing overheads, i.e. repartitioning computation time and migration, have to be minimized, diffusive methods provide substantial benefits.
Details
Original language | English |
---|---|
Pages | 154-157 |
Number of pages | 4 |
Publication status | Published - 2016 |
Peer-reviewed | Yes |
Conference
Title | EuroMPI 2016: The 23rd European MPI Users' Group Meeting |
---|---|
Conference number | |
Duration | 25 - 28 September 2016 |
Location | |
City | Edinburgh |
Country | United Kingdom |
External IDs
Scopus | 84995588305 |
---|---|
ORCID | /0000-0003-3137-0648/work/142238847 |
Keywords
Keywords
- dynamic load balancing, Workload Diffusion, HPC