الفهرس 
Chapter 1: IntroductionOnly 14 pages are availabe for public view
 |  | from | 18 |  |  |
| | | from | 18 | | |
Abstract
The exponential growth in computing resources has made distributed systems economically more attractive, for many applications, than the very expensive massively parallel machines. Distributed computing sys-tem (DCS) can be used to process a large job in a cooperative fashion. Where, the job has to be divided into smaller units, called tasks, and these tasks have to be distributed to appropriate computational ele-ments, to be concurrently processed. In theory, this approach can dra-matically improve the performance since each task can be executed on an architecture that is best suited for it. In reality, however, exploiting the full potential of the system requires efficient distribution of the tasks comprising the application onto the available computers in the system. If the distribution is not carefully implemented, the performance of the system may suffer. Computers in the system may spend most of their time waiting for each other instead of performing useful computations.
This thesis presents framework, called Map Reduce Agent Mobility (MRAM), for multitask scheduling in DCS to improve performance, reli-ability of distributed systems. The MRAM is developed based on mobile agent under Java Agent DEvelopment Framework (JADE). The perform-ance of the DCS is improved by using mobile agent in distributing the application tasks while the DCS reliability is improved by taking into ac-count both link failure and node failure during task distribution. Also, this thesis presents a modified framework called Enhanced MapReduce Agent Mobility (EMRAM) to support Big Data analysis on DCS environ-ment. The EMRAM is similar to the proposed MRAM with additional modifications developed in three different phases to overcome the drawbacks of Hadoop including task replication, centralized node and node failure.