Microsoft Word - 06-B-_인도-2권-지불예정-Copyright Accepted_ 825-fianl paper-new1.doc

Summary A Parallel system is collection of tightly coupled processors typically of the same type. In the present study, loosely coupled personal computers in a workgroup over the Intranet are going to be used. Though, networked machines are having different types of processors with varying clock speed yet computing on the networked machines are becoming very popular to solve both data intensive and compute intensive scientific problems due to the demand for higher performance and lower cost. Usually computational intensive areas have been referred to as scientific processing viz. linear algebra, information retrieval etc. One of the driving forces behind this shift is the availability of portable robust software to utilize and manage a network of PC's. As more and more organizations have high-speed local area networks interconnecting many general-purpose desktop PC's, the combined computational resources may exceed the power of a single high performance parallel computer. Such a design of distributed memory architecture using message passing interface between cooperative tasks within a parallel application is also called parallel system with loosely coupled processors. Since different tools and techniques are available to utilize and manage a network of PCs. It is believed that an efficient and affordable model of distributed parallel system based on the networked machines can be created with the use of these portable robust softwares and standardization of distributed parallel computing on network of PCs can be done. In this paper, we studied the behavior of parametric TCP/IP and optimize the communication among the networked machines keeping in view latency and communication load on bandwidth