Improving performance of benchmark method in evaluating storage system

Aim. An access feature extraction system is designed to record the workload in a real storage system, thus improving the evaluation performance of the benchmark method. Section 1 of the full paper explains how we implement this improvement with the help of Figs. 1 and 2. Its core consists of: each I/O of the storage system has many property values such as access time, start sector, request length etc; each I/O consists of the cache-matching step, the queue-merging step, the I/O scheduling step etc; the access feature extraction system arranges an accessing probe into every step of the I/O processing and records all property values of each I/O of the storage system in a real application; according to the I/O property values extracted, the system can produce the same workloads as in the real application. Section 2 evaluates the performance of the access feature extraction system by running a series of workloads. The experimental results, presented in Figs. 3 and 4, and their analysis show preliminarily that the performance of benchmark method in evaluating the storage system is improved, that the I/O property values recorded by the access feature extraction system are accurate and usable, and that the access feature extraction system has a minimal impact on the performance of the storage system.