Characteristics Analysis of Aggregate Workload on Storage System

In this paper we analyze aggregate workload characteristics of multiple I/O streams in shared storage environment such as storage area network (SAN). In these shared storage environments, independent and separate user data spaces make the access streams interleaved and localized. To evaluate overall storage performance or optimize data layout, it is important to understand aggregate workload characteristics of these streams which mutually influence and affect the overall system performance. We analyze access pattern and arrival pattern of aggregate workload by mathematics methods for the first time. The Poisson model is adopted for single stream since it is concise and straightforward. We validated that the mathematic analysis is accurate enough under our assumptions by simulation experiments using synthetic workload generator.