Name
Title | ||
|---|---|---|
Demonstrating improved application performance using dynamic monitoring and task mapping |
Abstract | ||
|---|---|---|
This work demonstrates the integration of monitoring, analysis, and feedback to perform application-to-resource mapping that adapts to both static architecture features and dynamic resource state. In particular, we present a framework for mapping MPI tasks to compute resources based on run-time analysis of system-wide network data, architecture-specific routing algorithms, and application communication patterns. We address several challenges. Within each node, we collect local utilization data. We consolidate that information to form a global view of system performance, accounting for system-wide factors including competing applications. We provide an interface for applications to query the global information. Then we exploit the system information to change the mapping of tasks to nodes so that system bottlenecks are avoided. We demonstrate the benefit of this monitoring and feedback by remapping MPI tasks based on route-length, bandwidth, and credit-stalls metrics for a parallel sparse matrix-vector multiplication kernel. In the best case, remapping based on dynamic network information in a congested environment recovered 48.9% of the time lost to congestion, reducing matrix-vector multiplication time by 7.8%. Our experiments focus on the Cray XE/XK platform, but the integration concepts are generally applicable to any platform for which applicable metrics and route knowledge can be obtained. |
Year | DOI | Venue |
|---|---|---|
2014 | 10.1109/CLUSTER.2014.6968670 | Cluster Computing |
Keywords | Field | DocType |
application program interfaces,matrix algebra,message passing,parallel processing,task analysis,vectors,Cray XE/XK platform,MPI tasks,application communication patterns,application performance,application-to-resource mapping,architecture-specific routing algorithms,credit-stalls metrics,dynamic monitoring,feedback,parallel sparse matrix-vector multiplication kernel,run-time analysis,system-wide network data,task mapping | Dynamic network analysis,Kernel (linear algebra),Architecture,Computer science,Task mapping,Parallel computing,Real-time computing,Exploit,Multiplication,Bandwidth (signal processing),Dynamic priority scheduling,Distributed computing | Conference |
ISSN | Citations | PageRank |
1552-5244 | 3 | 0.44 |
References | Authors | |
13 | 4 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jim M. Brandt | 1 | 70 | 10.20 |
| Karen D. Devine | 2 | 404 | 24.66 |
| Ann C. Gentile | 3 | 37 | 7.91 |
| Kevin T. Pedretti | 4 | 196 | 21.20 |