Resource discovery in Grid systems based on the Distributed Hash Table

摘要

Resource discovery is one of the key functionalities of the Gridudsystems, since it is used by a variety of different services. The mainudgoal of every resource discovery system is to collect the data aboutudall the Grid resources and to serve those data to all the users andudapplications that are interested in them. The usage of the resourceuddiscovery service is simple: we provide it a search query as an inputudand it returns us the list of locations of all the resources whichudsatisfy all the conditions provided by the query.ududThis work presents a distributed system that is used for resourceuddiscovery in Grid systems. The system is based on structuredudpeer-to-peer technologies; more precisely, on the Distributed HashudTable. It enables us to search for the Grid resources usingudexact-match queries as well as range queries. The support for rangeudqueries is very important in Grid systems since those queries are theudmost frequently used. The weakness of the majority of structuredudpeer-to-peer resource discovery systems is the inefficientudimplementation of the support for range queries. In our system, on theudother hand, the amount of network traffic that is generated during theudresolution of range queries is the same as with exact-match queries. ududAdditionally, the support for multi-attribute queries (i.e., theudqueries that contain multiple search conditions) is also moreudefficient since the amount of network traffic when resolvingudmulti-attribute queries is also lower than with the otherudresource-discovery systems. ududDue to both the properties mentioned above and due to the betterudresilience to node failures, our system has better scalability thanudthe rest of the resource-discovery systems that are based on audstructured peer-to-peer system. As a consequence, its performance inudlarge Grid systems is better than the other systems. ud