Abstract | ||
---|---|---|
Processor resources required for an effective execution of anapplication vary across different sections. We propose to takeadvantage of clustering to turn-off resources that do notcontribute to improve performance. First, we present a simplehardware scheme to dynamically compute the energy consumed by eachprocessor block and the energy-delay2 product for a given intervalof time. This scheme is used to compute the effectiveness of thecurrent configuration in terms of energy-delay2 and evaluate thebenefits of increasing/decreasing the number of active issuequeues. Performance evaluation shows an average energy-delay2product improvement of 18%, and up to 50% for some applications, ina quad-cluster architecture. |
Year | DOI | Venue |
---|---|---|
2003 | 10.1109/ICCD.2003.1240922 | ICCD |
Keywords | Field | DocType |
dynamic cluster resizing,ina quad-cluster architecture,energy-delay2 product,simplehardware scheme,effective execution,different section,intervalof time,eachprocessor block,performance evaluation,active issuequeues,energy-delay2product improvement,statistical analysis,queueing theory,computer architecture,resource allocation | Resizing,Computer science,Queue,Parallel computing,Real-time computing,Queueing theory,Resource allocation,Cluster analysis,Statistical analysis | Conference |
ISSN | ISBN | Citations |
1063-6404 | 0-7695-2025-1 | 5 |
PageRank | References | Authors |
0.86 | 8 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
José González | 1 | 526 | 35.85 |
Antonio González | 2 | 3178 | 229.66 |