Performance Evaluation and Monitoring
ACM Computing Surveys (CSUR)
On the construction of a representative synthetic workload
Communications of the ACM
Synthetic program specifications for performance evaluation
ACM '72 Proceedings of the ACM annual conference - Volume 2
Two approaches for measuring the performance of time-sharing systems
SOSP '69 Proceedings of the second symposium on Operating systems principles
An experiment in the use of synthetic programs for system benchmarking
AFIPS '74 Proceedings of the May 6-10, 1974, national computer conference and exposition
| Hi-index | 0.02 |
THE MOST IMPORTANT QUESTION in any evaluation of a computer should be, “How long will it take this system to process my workload?” If the system cannot do the workload, there is no point in asking how much it costs or any other facts about the system. With this thought in mind, it then becomes apparent why the technique used to determine the answer to this “how long” question becomes the most important item in any computer evaluation. Time does not permit a detailed discussion of both: why application benchmarks are the best of the timing techniques available, and of the method for setting up application benchmarks which will be truly representative of the applications to be run. Therefore, the second topic (Setting Up Application Benchmarks) has been chosen, because I believe most who have tried other timing techniques know there must be something better. I also feel that those who have used application benchmarks have limited their thinking to sequential processing systems and should now consider the preparation of application benchmarks for systems capable of multi-processing and/or multi-programming.