A New Approach to I/0 Performance Evaluatio-n- — Self-Scaling I/0 Benchmarks, Predicted I/0 Performance

Current 1/0 benchmarks suffer from several chronic problems: they quickly become obsolete; they do not stress the 1/0 system; and they do not help much in understanding 1/0 system performance. We propose a new approach to 1/0 performance analysis. First, we propose a self-scaling benchmark that dynamically adjusts aspects of its workload according to the perfor-mance characteristic of the system being measured, By doing so, the benchmark automatically scales across current and future systems. The evaluation aids in understanding system perfor-mance by reporting how performance varies according to each of five workload parameters. Second, we propose predicted performance, a technique for using the results from the self-scaling evaluation to estimate quickly the performance for workloads that have not been measured. We show that this technique yields reasonably accurate performance estimates and argue that this method gives a far more accurate comparative performance evaluation than traditional single-point benchmarks. We apply our new evaluation technique by measuring a SPARCstation 1 + with one SCSI disk, an HP 730 with one SCSI-II disk, a DECstation 5000/200 running the Sprite LFS operating system with a three-disk disk array, a Convex C240 minisupercompute