|
Title:
|
OBJECT CLUSTER COMPUTING. ADVANTAGES OF OBJECT ORIENTED LOAD BALANCING STRATEGIES |
|
Author(s):
|
J. A. Álvarez , J. Roca , J. C. Ortega , I. García |
|
ISBN:
|
972-98947-3-6 |
|
Editors:
|
Nuno Guimarães and Pedro Isaías |
|
Year:
|
2004 |
|
Edition:
|
Single |
|
Keywords:
|
Cluster, object, parallel programming, load imbalance. |
|
Type:
|
Short Paper |
|
First Page:
|
2147 |
|
Last Page:
|
2150 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Nowadays, clusters of workstations have shown to be a cost-effective solution for parallel computing. However, the performance of the applications in these environments may be dramatically affected by external factors involving load imbalance. Object orientation is a programming paradigm that has proved to be more effective than the traditional stand alone process programming. The use of objects in the development of parallel implementations for real world applications makes load balancing easier. The selection of a specific load balancing strategy depends on the target application and on the underlying hardware platform. This paper describes the basis and benefits of using object orientation and parallelism together and our ongoing research work on dynamic load balancing strategies. It also describes the advantages of using such objects based strategies which preserves data locality in migration operations. Their behavior is also studied. The implementation of these strategies makes use of the Charm++ abstraction layer
(C++ based framework) from Charm framework (based on objects and message driven at execution). |
|
|
|
|
|
|
