Design and analysis of high-performance and recoverable data storages

With explosive growth of networked information services and e-commerce, data protection and recovery have become the top priority of business organizations and government institutions. In order to make data recoverable and provide high-performance storage service, we have conducted intensive research on networked storage systems. This dissertation addresses the issues related to high recoverable and high-performance storage systems.;First, this thesis presents an implementation and performance evaluation of two snapshot technologies: copy-on-write and redirect-on-write snapshots. Our experimental results uncovered many interesting observations regarding the performance characteristics of the two snapshot techniques. Depending on the applications and different I/O workloads, the two snapshot techniques perform quit differently.;Second, this thesis proposes a new disk array architecture for continuous data protection. This architecture provides Timely Recovery to Any Point-in-time named TRAP-Array. We have implemented a prototype TRAP architecture using software at the block device level. Our experiments demonstrated that TRAP is not only able to recover data to any point-in-time very quickly upon a failure but it also uses less storage space than traditional daily incremental backup/snapshot. Compared to the state-of-the-art continuous data protection technologies, TRAP saves disk storage space by one to two orders of magnitude with a simple and a fast encoding algorithm.;Third, this thesis presents a theoretical study on COW snapshots and incremental backups. Our theoretical work uncovered the fundamental limitations of the existing data protection technologies and provides the first theoretical explanation as to why so many data recoveries (over 67% recoveries) failed using these existing technologies. Based on our theoretical results, we propose a new storage architecture that overcomes the limitations of existing technologies. Our experiments established recoverability and performance of the new data protection technology as well as the existing ones.;Fourth, this thesis presents a design and implementation of Time-frequency feature extraction from Electromyography (EMG) based on the Graphic Processing Unit. The results indicated that GPU significantly increased the computation speed up to 50 times faster than the CPU setting. High performance GPU shows a great promise for EMG-controlled artificial legs and other applications that need high-speed and real-time computation.