An alternative ranking scheme for computational grid resource selection
With widespread use of Internet and other communication technologies, it has become extremely popular to employ idle processing power of computers, which distributed around the world to achieve high performance computing. Grid computing is one the solutions to compute and process large amount of dat...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/37025/5/MehranRanjbarMFSKSM2013.pdf http://eprints.utm.my/id/eprint/37025/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:70093?site_name=Restricted Repository |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | With widespread use of Internet and other communication technologies, it has become extremely popular to employ idle processing power of computers, which distributed around the world to achieve high performance computing. Grid computing is one the solutions to compute and process large amount of data by using distributed systems. Find suitable resources among numerous resources to execute the submitted job is a critical problem in grid environment. The main idea of proposed algorithm for resource selection is finding reliable resources that can meet job requirements and then based on available resources characteristics and processing power decides about workload of each subtasks. The number and workload of each subtask directly related to the number of available resources and their specifications such as CPU, memory and the status of the network that they are connected. In addition, a ranking mechanism developed to assign a rank to all resources, which join the selection process. It assists reliable resources have more chance to be selected among resources with similar processing performance. Experimental results present that proposed algorithm achieved maximum 95.4% success rate among 1000 submitted job which is 3.3% better than current success rate achieved by Grid Resource Vector (GRV) . The proposed algorithm achieved maximum one second execution time which is showing 2.7s improvement among evaluated methods. In addition, it showed 251.12s completion time among a job with 100K length with 5000 resources where evaluated method achieved 376.45s that shows 125.33s improvement. Proposed algorithm succeeds to increase execution success rate and improve average execution and completion time in comparison of similar algorithms, which shows that alternative ranking scheme is effective for selection reliable resources. |
|---|