Research on Self-Adaptive Distributed Storage System

Ensuring performance is a long term goal in network storage systems. Existing management tools require detailed knowledge of the system characteristics and the access patterns to modulate a suitable resource schedule, but the knowledge is hard for system administrator to get. This paper brings out a self-adaptive distributed storage network (SDSS). By using an online QoS monitor controller, the SDSS can adaptively allocate storage bandwidth to application classes so as to meet their response time requirements. Firstly, in the SDSS, spare bandwidth from underloaded classes should be reallocated to classes that are heavily loaded; Secondly, when aggregate workload exceeds SDSS capacity, causing the target response times of all classes to be exceeded, SDSS allocates a wider bandwidth for applications that has higher priority; Thirdly, the SDSS could online automatically change the data distributions when the storage system is heavy loading or overloading, therefore it can reduce the I/O burden of hot spots, and guarantee QoS of storage systems. Our experiments demonstrate the improvement of the storage systems by using SDSS.