الفهرس 
الغلافOnly 14 pages are availabe for public view
 |  | from | 227 |  |  |
| | | from | 227 | | |
Abstract
The Grid is an emerging technology that aims to enable resource sharing and coordinated problem solving in dynamic multi-institutional heterogeneous virtual organizations.
The Grid computing objective is to create the illusion of a simple but large and powerful self-managing virtual computer. This application will be emerging out of a large collection of connected heterogeneous systems that share various combinations• of resources.
Resource management is one of the key challenging problems of Grid computing. It schedules applications, over the available resources. Such scheduling, over a large scale distributed environment, raises performance issues.
A survey of current implementations of Grid Resource Management like Globus, Legion, Nimrod-G showed that they applied different mechanisms for keeping Grid resources’ information and they used different protocols to keep Grid resources’ information updated, for instance, push protocol which is applied in Globus project. This protocol periodically updates the resource information. There is also ’condocstartd’ dameon which periodically advertise current resource information in Condor project. However most of those systems paid little consideration to system performance, as they ignored issues like resource availability and workload, amid regularly advertising the resource infonnation. Finally such protocols caused noticeable performance headache in such implementations.
Agent-based Resource Management System (ARMS) is one of the implementations of Grid resource management systems that addresses Grid key challenges, such as scalability and adaptability.
In addition it monitors performance of the system and applied different performance optimization strategies suitable for the current system’s performance status.
However the current performance optimization strategies still suffer from some performance troubles. Such troubles are improved in this work with an adaptable decentralized resource information dissemination strategy, to solve the performance issues, that occur due to service advertisement and discovery processes within ARMS system.
In this work, Performance Monitoring and Advisory (PMA) for ARMS component, is capable of performance modeling and simulation in ARMS to improve the overall system performance. It is implemented to compare the new performance optimization strategy with those previously defined in literature.
Several experiments carried out with different environmental conditions to simulate the real-world Grid scenarios, recorded noticeable better performance for the new strategy. This is especially due to the fact that it is more adaptable to continuous changes in both workload and structure of ARMS system.
Keywords: Grid computing, Resource management, performance optimization, Decentralized service advertisement, Agent-based resource management.