Study On Key Technologies Of P2P Based Resource Management In Grid

Grid is an emerging infrastructure which enables effective coordinate access to various distributed computing resources in order to serve the needs of collaborative research and work across the world. Grid resource management is always a key subject in the grid computing. There are two main problems: how to offer more flexible and intelligent grid resource management service with its original knowledge and how to enhance the scalability of grid resource management system so that it could adapt to the expanding grid scale.In order to resolve the above problems, this dissertation begins its study of grid resource management technology with the high scalable model construct and knowledge resource management. Making use of semantic web technology and DHT-based overlay network technology, creative achievements in knowledge representation, knowledge storage, resource discovery and schedule are as follows:1. A grid resource management model, KRMM, is proposed. In this model, resources are managed according to the local-global grid resource knowledge bases. LKB (Local Knowledge Base) manages the resource of local domain autonomically while the distributed GKB (Global Knowledge Base) realizes the coordination and share of global scale grid resources. This model not only satisfies the need of grid resource autonomy and effective management, but guarantees the scalability of the whole grid system and avoids single point failure.2. A distributed storage model of global grid knowledge base, which is based on DHT overlay network, is proposed. Resource servers of local domains connect each other according to DHT overlay network, disassemble the global database and store knowledge data in a distributed way. This storage model is high scalable and self-adaptive. Corresponding algorithm is put forward in order to solve the problem of building and maintaining such a global knowledge database of resources in the grid.3. A common ontology of KRMM grid is established, and an integrated mechanism of distributed resource ontology is also proposed. In KRMM, each site of grid could keep partial gird resource ontology. The common resource ontology is a template on which local resource ontology is developed in each site and provides interoperability among each site. The global ontology of the grid is integrated by local resource ontology of each site. And with the support of the global ontology, each grid site which keeping only segmental ontology could find semantic-matched resource from the global grid.4. A knowledge-based resource and request description approach is proposed. This approach can describe grid resource aggregation in a network-centric manner. And a knowledge-based resource cooperating discovery algorithm is designed and proposed.5. A grid resource scheduling algorithm across multi-sites based on decentralized scheduling model is proposed. This algorithm, centered on Backfilling strategy, discusses the problem of grid resources scheduling across multi-sites under the condition of heterogeneity of processor amount.