Dynamic Replication And Resource Management In Wide-area Storage Virtualization System

More data-intensive and high-performance computing applications have appeared that require the efficient management and transfer of terabytes or petabytes of information in wide-area, distributed computing environments. But the traditional storage devices can't satisfy the demand of those applications. That requires a storage visualization system that provides a uniform interface for connecting to heterogeneous data resources over a network and accessing replicated data sets. Such is the Global Distributed Storage System (GDSS).Dynamic replication and storage resource management system are two essential parts of GDSS. Dynamic replication system can add replicas of source files in a domain that can improve the load balance and availability. Otherwise, replicas can also be produced in a remote domain that can decrease the user access latency and bandwidth consume. Replication system of GDSS includes five modules: replica catalog, access logger, resource manager, replica manager and replica transporter etc. The system can create replicas according to the user access pattern, but also delete the leave unused replicas to make full use of the storage space. As addition, the system provide a parallel transferring strategies among several replicas that gets different data segments from replicas and joint the segments at the client. The parallel transferring strategies can improve the data transferring efficiency.Replica location strategies and replica consistency are two key issues in GDSS. Replica location strategy mainly decides how and where to create replicas on the base of data access cost. The latter focuses on how to guarantee the consistency among the replicas in the WAN.GDSS replication model combines FastSpread and Cascading strategies to provide a T-value replica expand strategies that is a good choice of replica location strategies in the storage visualization environment.GDSS storage management model is responsible for arranging, location, registering, scheduling, allocation and asserting of storage resource. Storage management model can view different storage medias such as HTTP, FTP, SAN,NAS etc. as a same storage node and allocate these storage nodes to user.At the end, GDSS has a good management and monitor web interface. Through the interface, system administrator can configure the system , strategies , algorithm parameter and monitor the status of users access and storage agents.