Framework for Resource Management in A Grid Environment

Abstract

Grid Computing heralds the dawn of a new paradigm for next-generation high performance computing. As a descendent of distributed computing, it enables selection, sharing and aggregation of geographically distributed resources for solving intricate problems of science, engineering and commerce, which have phenomenal computational needs. It offers reasonable, consistent and pervasive access to high-end computation, mainly as a result of availability of faster hardware, more sophisticated software and escalation of Internet. The principles of Grid computing focus on large-scale resource sharing in distributed systems in a flexible, secure, and coordinated fashion. Resource management forms the core of Grid computing. Resources in a Grid are often geographically distributed, heterogeneous and dynamic in nature. Resources are owned by different organizations with different usage policies, cost models, varying loads and availability patterns. The conventional resource management schemes are based on the assumption that resources are owned by a single organization and therefore, would be available exclusively. There are relatively static models that have centralized controller that manage jobs and resources accordingly. These assumptions do not hold true for a Grid environment. Therefore, in spite of a number of advances in Grid Computing, resource management persists to be a challenging, complex, yet most significant task and has been the main focus of this research work. Initially, a detailed review of the work done in the area of Grid Resource Management has been done. It analyzes, compares and reports existing approaches to Grid resource management, state of the art in Grid resource management and the challenges, that need to be addressed. A comparative analysis of middleware technologies has been done for Grid resource management and fault management mechanisms. The existing Grid resource management systems have also been explored for their resource discovery, resource monitoring, job execution, provisioning, brokering and scheduling scenarios. The fault management approaches in existing middleware and existing monitoring systems have been explored, compared and reported. Related technologies like Semantic Web Technologies and Autonomic Computing have also been studied and reported. Based on the literature survey, it is apparent that issues of interoperability, provisioning and fault management are the main challenges besides numerous other issues that need to be addressed. To address these diverse Grid resource management challenges, a resource management framework, named PRATHAM has been initially proposed and further designed, developed and tested in this work. The proposed Grid Resource Management framework, PRATHAM inspires from Semantic Web Technologies and Autonomic Computing. It has a five-layer architecture, with Fabric layer as the base layer which constitutes the resources; second layer is Grid Middleware, and it comprises of the middleware. Next is Broker layer, which renders self-management to the framework; fourth, the Information layer, realizes the interoperability and provisioning capabilities of the Grid. The top-most fifth layer is the Portal, which provides an interface to the Grid. PRATHAM provides a hierarchical model based distributed framework and it works well for large-scale environments. It is an endeavor to provide Grid users not only the freedom of higher operability but also enhanced resource provisioning within a Grid environment, along with better monitoring and self-management facilities. Interoperability has been addressed through semantics; user requirements have been optimized and provisioned by providing facility of advanced reservation of resources to the user. Apart from this, it provides a monitoring system based on pulse monitoring which has been implemented through ganglia and helps in monitoring individual resources and status of resource queues; self-management has been realized through Autonomic Computing principles. This research work implements PRATHAM through Globus Toolkit 4 and Alchemi.NET besides various other tools. The framework has been deployed at Thapar University, Patiala. Experiments conducted clearly demonstrate its usage and various features. Finally, the framework has been compared with existing Resource Management frameworks and Resource Management systems to validate the outcomes. The results show that PRATHAM successfully and collectively addresses the issues of interoperability; resource provisioning and self-management, the OGSA requirements for future Grids.