Basser Seminar Series

Postgraduate Research Seminar Series

Grid Computing: Challenges and Opportunities

Mohammed Aal Salem
Networks and Systems Research Laboratory

Wednesday 18 May 2005, 2pm

Basser Conference Room, Madsen Building G92

Abstract

Distributed computing is a topic that has been studied a lot in earlier periods, but in recent times a new field, Grid computing, has brought attention back to the area. Several works on grid computing have been anticipated in the last few years. However, Grid environments pose many challenges, including those of security, the application itself, Fault Tolerance, Quality of Services (QoS) and Grid Protocols. The Grid computing platform consists of several sites that are interconnected over the internet. These sites can be linked and classified as. Firstly, Intra-site "Intragrid" which, links computers or resources in a local area. With these links, Bandwidth is shared among all the users and is the same across all transfers. Secondly, Inter-site "Intergrid", which links sites using the internet as a backbone medium of transfer or communication and relying on it is huge physical bandwidth. On these links, there is no sharing of bandwidth and it depends on the internet characteristics and needs the QoS to support this interconnection A Grid computing system uses the high performance computing (HPC) which depends on ATM technology to establish a connection between several sites. In addition, this leads to problems with bandwidth sharing and latency. Also it needs very high concern by QoS. This thesis was primarily involved in investigation the areas of network latency, bandwidth sharing, number of hosts, speed of hosts and number of tasks in the Intragrid. The experimental results show how can improve the high performance computing.