Research On Metadata Server Faced Data Allocation Policy

With the explosive growth of information, storage industry becomes a booming point in IT industry. The exponential increase of data and ever changing storage demands of modern applications bring new challenge to the area of storage management. Storage management is critical to the availability, scalability and high performance of the storage system.In a Storage Area Network (SAN) which consists of client, metadata server and multiple storage nodes, metadata server is responsible for the management of storage system and plays an important role. In a block-interfaced distributed storage system, the design of managing metadata at block-level is presented by using storage virtualization techniques and data allocation techniques. A smaller granularity of data management compared to file-level management is provided and the demands of devices, balancing workloads of the storage system and improving the performance of the system are considered. Storage resources allocation and address mapping polices are critical to the performance of the storage system, so three kinds of address mapping polices are designed and analyzed including Static Address Mapping, Dynamic Address Mapping and Improved Dynamic Address Mapping. The appropriate approach can be chosen to meet certain needs of applications, therefore disk workloads are balanced and storage system performance is improved.The metadata server is implemented in a Virtual Interface (VI) Protocol based storage area network. The storage system runs on Red Hat Linux 9.0 (Kernel Version: 2.4.20) and developed with C language. The evaluation results demonstrate that metadata server faced data allocation policy is efficient and flexible.