SPC1-Read workload

SPC-1 ([36,17]) is a widely used commercial benchmark provided by the Storage Performance Council. It uses a sophisticated workload that simulates business critical environments like OLTP systems, database systems and mail server applications. We use a prototype implementation of the SPC-1 benchmark called SPC1-Read, that matches the specifications of the read component of the SPC1 workload([37]), which comprises uniform random (10%), hierarchical reuse random (65%), and incremental sequential (25%) access patterns. The $thinktime$ and $readsize$ of the constituent workloads are continuously varied according to the probability distribution prescribed by the benchmark. The impact that the write component has on concurrent reads depends on the choice of the size of the write cache, the order of destages as well as the timing of the destages([38]). It serves us well to ignore the write component of the workload so as to clearly appreciate the relative performance of the prefetching algorithms without diluting the results with writes.