Research On The Key Aspects Of Grid-Based Storage System

Data, as the core resource in the information era, have been distributed to the networks and stored in the networks. Currently, the Grid-based distributed fault storage (Grid Storage), as a new research field, is one of the services provided by Grid. Grid Storage is different from Grid Computing. Grid Computing is to scavenge the idle cycles, while Grid Storage is to aggregate the idle and voluntary storage resources. Grid Storage, as the initiative research, has not deeply been studied by the researchers and a fully practical Grid Storage solution has not been gotten. The Grid Storage research has some difficulty and is very important both in theory and applications.In the dissertation, the key aspects of Grid Storage are systematically and deeply studied. The aggregation of node storage resources in the Grid Storage system, multi-region content distribution under the underlying Peer to Peer networks, storage resources sharing and data distributed fault tolerant storage in the Grid Storage system are mainly studied.Firstly, we survey the history, development, applications of the Grid systems, and the research background, current research state and active research field are also described. In Chapter Two, Grid Computing and Peer to Peer Computing are compared in detail. Chapter Three studies how to aggregate Grid node storage resources, including node organization mechanism, node join and leaving, storage resource scheduling and allocation, resource node searching, node unallocated storage space management, and storage resource allocation and reclaim. And the stochastic fluid model is adopted to analyze the performance of the whole Grid Storage system. Chapter Four studies the data distribution mechanism among multi-region Peer to Peer network which is the underlying of Grid Storage system. Chapter Five studies the sharing problem of Grid Storage resources, including the sharing model, mathematical description, and the optimal function. Finally the system load and respond time are simulated and analyzed. In Chapter Six, the distributed fault tolerant storage is studied due to node dynamic join and leaving. Based on the work of RAID, coding theory and Peer to Peer network fault tolerant storage, we further study the General Weighted Distributed Hash Table objection function by which data can be fairly distributed to the storage nodes. Then we study the data fault tolerant architecture with combination of LT code and General Weighted Distributed Hash Tables.The research on Grid Storage involves the knowledge of Grid Computing, Peer to Peer network, Control Theory, Operations Research, Coding Theory, and Fault Tolerance. The contributions of the dissertation are summarized as follows:(1) The characteristics of Grid Computing and Peer to Peer Computing, including the concept, goals, architecture, classification, organization, resources, security, applications, infrastructures, characters, etc., are all analyzed in detail.(2) The campus-based scientific Grid cooperation environment is proposed to support the collaboration of multi-discipline in the university. For the large organizations(such as university, large company with many sub-departments,...