Abstract:
Grid computing provides sharing of heterogeneous and distributed resources connected through a network in a collaborative environment. So there is a need of a service in such kind of environment which will discover, allocate, monitor and manage the use of network accessible capabilities in order to achieve quality of service in a completely transparent way. Such services are provided by the Resource Broker which is a key entity and requirement in every Grid computing infrastructure. Current implementations of Brokering service do not provide the required ease and comfort and the required level of transparency to the end user. Scheduling process in some systems is very complex and they do not handle the dynamic and variant behavior of distributed resources efficiently, and scheduling decision made by the Broker is unrealistic in most of the cases. In this project we will present a broader and efficient design architecture and framework for a Resource Broker form the perspective of general application needs which handles the dynamic, unpredictable and ever changing behavior of distributed computational resources in the Grid. Our proposed architecture extends the brokering and scheduling framework by taking it form middleware to the level from where it could be easily integrated with the Operating System kernel. This will give it more control over underlying computing resources such as CPU, memory and bandwidth and also increase the efficiency and transparency with respect to the user. We will also describe a realistic approach taken by the brokering engine in making its decision about selection of optimum resources across the Grid. The interaction of the Resource Broker with other services and components of Grid computing infrastructure will also be described.
