Research On Intelligent Process Methods Of Object-based Network Storage

The amount of data in many aspects such as scientific computing and E-Business keeps on exploding. Physical storage devices and controlling components have been separated from traditional computing systems and evolved a scalable, intelligent storage subsystem, which should provide impressive storage interface, data allocation, storage management, backup and archive, and so on. The storage system is required for high performance, high scalability, high availability, high reliability and high security, etc. Advanced research has been propagated in this field and remains challengeable.To satisfy such mounting demand, the intelligence of storage system is studied in detail in this dissertation. The main contents are as follows: the research on the scalable object-based storage system, storage active service, the adaptive multi-parallel pipelining I/O scheduling, I/O path, the intelligentized hierarchical storage and the verification of object-based file system based on linear temporal logic.In scalable object-based storage system (called HOSS), Object Storage Controller (OSC) has some intelligent features, by extending the definition of object of ANSI T10 Technical Committee. In the HOSS, the policy for object allocation is a critical aspect affecting the overall systems performance. An efficient Hybrid-Layout algorithm for static load balancing is provided. The goals of the storage active service are to combine and distill previous work into a simple and usable model, design a storage active service mechanism as an understandable extension to object, and most importantly build a working system - a system that augments conventional object management with data, attributes and trigger as an integral part of the storage object. At the same time, numerous motivating practices for storage active service and lays out preliminary plans for the forecast mechanism based on psychology are provided. Based on Region Peer to Region Peer (RP2RP), a dynamic load balancing policy is discussed. Also, simulation results show that the method of storage active service is practicable.Application workload is decompounded some sub-workloads for parallel pipelining. In the same pipelining stage (composed of function component or device), the number of sub-workloads is called pipelining parallel degree, these sub-workloads produce synchronal overhead. The I/O path in network storage and segmenting the reasonable pipelining stage are studied, and I/O pipelining mechanism is provided which may have significant direction for enhancing performance of object-based storage system. Additional experiments show that the adaptive I/O scheduling implemented by multi parallel degree pipelining can well overlap I/O and computation in HOSS.Built on a sequential write/read device, a tape library is seldom considered as a viable place for fast backup/restore data. However, for hiberarchy storage, with the help of the object-based storage and storage virtualization technologies, a backup and archive system based on Object Digest and a tape virtualization system are impelemented in accordance with total performance/cost of ownership. The tape virtualization system has the automatic data transfer algorithm based on Information Lifecycle Management (ILM). The prototype system performance is presented and analyzed to achieve higher read/write performance.An object file system temporal logic, called OFSTL, is a temporal logic proposed for the specification of object-based file systems. It is an extension of Manna and Pnueli's linear temporal logic. It can express both the properties and the implementations of object-based file systems. OFSTL makes it feasible to understand object-based file system access behavior and extract object-based file system access semantics.