Storage subsystem workload characterization in a real life personal computer environment

Storage subsystem performance is a critical factor in the overall computer system performance. To evaluate and improve the performance of existing and future storage subsystems for an environment, the workload as seen by the storage subsystem in that environment has to be characterized. In the paper, storage subsystem workload in a real life Netware personal computer environment is traced using the storage subsystem tracing tool (SSTT) introduced in Ali et al. (1992). SSTT is an operating system independent tool which is shown to introduce minimal overhead. Workload characterization is then performed by using analysis tools that have been developed to study important statistical parameters of the traced workload such as logical block address (LBA) frequency, IO request interarrival time frequency, IO request size, and the instantaneous demand on the storage subsystem. Distributions for these important parameters are obtained and analyzed.