Skip to Main Content
It is widely believed that future computing environment will consist of logically connected but geographically distributed computation and storage resources with a variety of capabilities. Furthermore, these systems typically go through rather complicated upgrades which lead to heterogeneity over time. Meanwhile, there are various applications that exhibit different I/O access patterns. In this paper, we proposed a novel scheme called service oriented storage (SOS) to automatically adapt diverse storage systems to their services in a large data grid environment. Using clustering algorithm based Pattern Discovery, Pattern Divisive, Storage Agent and Dissimilation Diagnosis modules, SOS attempts to identify proper storage nodes from the storage pool based on interactions of the user's I/O attribute. In other words, structure and attribute of the storage systems can adapt to the changed workloads and allow system service to achieve optimal and maximum self adaptively.