ABSTRACT
Symmetric multiprocessor systems (SMPs) have become a standard parallel processing platform for various applications. One important usage of such systems is the execution of commercial workloads, which represent one of the most rapidly growing market segments. In comparisons with scientific, numeric-intensive, and engineering applications, commercial workloads contain more sophisticated system software activities. Performance of commercial workloads is determined by so many factors (both hardware and software) that it is hard and time-consuming to evaluate. Based on an intensive experimental study conducted at Western Research Lab of Compaq, Barroso et al. [2] recently summarize the challenges and difficulties of the performance evaluation of commercial workloads — the lack of experimental resources. It is difficult and expensive to have a large scale and complex database engine running on a multiprocessor system with powerful I/O facilities for performance evaluation. In addition, the experiments for case studies have been shown to be highly time-consuming. So far, all published results on performance evaluation of commercial workloads on SMP multiprocessors are experimentally oriented and measurement-based case studies (see e.g., [7, 8, 10]).
