A proximity-aware semantic-based decentralized resource discovery framework for computational grids

Resource discovery is a service of Grid resource management that is considered as a core part of computational Grids. The service is one of the fundamental requirements of highly dynamic and heterogeneous computational Grids, which deals with providing appropriate resources for users to meet their jobs requirements. Existing research reveals that centralized and hierarchical resource discovery models can perform poorly for large-size Grids, because of various limitations. As a result, a decentralized resource discovery is recommended for large-size Grids. However, a lack of coordination between users and providers in a decentralized computational Grid environment often results in user jobs failing to find appropriate resources. One of the reasons for rejection of jobs is the usage of fixed schema between user requests and providers' availability that can affect the overall performance of the Grid. The resource discovery performance in decentralized Grid environment can be categorized based on four main drawbacks – high communication overheads, low job success probability, high latency and low resource utilization. To eliminate these drawbacks and improve performance, a proximity-aware semantic-based decentralized resource discovery framework is proposed for computational Grids. The decentralized resource discovery framework is developed and implemented incrementally using a combination of both semantic and proximity criteria. Extensive experimental analysis indicates that Pastry-based resource discovery model outperforms Chord-based one in terms of reducing communication overheads. To increase job success probability in decentralized resource discovery, the thesis extends current semantic techniques by presenting a sub-domain based ontological structure. Towards this end, a semantic-based decentralized resource discovery model is developed and implemented. For further enhancement of the applications performance and reduction of latency, the thesis proposes and designs the UPSARS (Unification of Proximity and Semantic for Appropriate Resource Selection) model for the sub-domain based semantic decentralized resource discovery. The experimental results of the UPSARS indicate that the model can reduce latency by allowing an application to allocate resources in proximity. Finally, to improve resource utilization, the thesis proposes a fuzzy-based framework for the selection of the most suitable resources by adding not only semantic and proximity factors, but also including the other parameters of the matched Grid resources such as number of machines and number of processors. Overall, the proximity-aware semantic-based decentralized resource discovery framework reduces communication overheads, enhances job success probability, reduces latency and improves resource utilization in a computational Grid environment