Study Of Load Balance Stragety In Object-Based Storage System

Object-Based Storage(OBS) is poised to be the next frontier in the network storage field, combining the benefits of the Network Attached Storage (NAS) and Storage Area Network(SAN) architectures. OBS interface standard is focus on moving chosen low-level storage, space management, and security functions into Object-Based Storage Devices(OSD) to enable the creation of scalable, self-managed, protected and heterogeneous shared storage for storage networks.Load balancing is a key issue in large-scale OBS system. This paper focuses on two steps. Firstly, the load balance strategy of OBS system is studied. Confirming the guideline for system's workload by analyzing the difference of OBS system and other distribute systems. Adopting self-adapting over loading strategy for finer performance of system's parallel data storage. Suggesting the notion of load balance degree for weighing the workload of OSDs. Adopting local distribute scheduling stragety and bidirectional driving strategy.Secondly, the objects in OSDs, which the basic storage units, are composed of data, attributes and methods. Synthesizing the attributes of objects in OSD, including granularity and accessing frequency, and I/O task's queue time, based on revenue/expenditure analytical method, a selection algorithm of load balance migrating objects is proposed. This algorithm is used to solve the even load problem of object storage system, making best use of attributes of data objects in OSD, not only reducing tasks'average response time, but also solve the access problem of hot objects. In addition, this method reduces the cost of migrating to the best of its ability. The experiment results suggest that this strategy can enhance the system's capability of parallel data storage extraordinarily, and reduce objects'I/O task average response time.