Quantitative system performance: computer system analysis using queueing network models
Quantitative system performance: computer system analysis using queueing network models
Analytic Queueing Network Models for Parallel Processing of Task Systems
IEEE Transactions on Computers
New computer methods for global optimization
New computer methods for global optimization
Operating system concepts (3rd ed.)
Operating system concepts (3rd ed.)
IEEE INFOCOM '92 Proceedings of the eleventh annual joint conference of the IEEE computer and communications societies on One world through communications (Vol. 3)
Performance bounds for concurrent software with rendezvous
Performance Evaluation
Mean-Value Analysis of Closed Multichain Queuing Networks
Journal of the ACM (JACM)
CASCON '93 Proceedings of the 1993 conference of the Centre for Advanced Studies on Collaborative research: distributed computing - Volume 2
Splitting techniques for interval parameters and their application to performance models
Performance Evaluation
Characterization and Analysis of Software and Computer Systems with Uncertainties and Variabilities
Performance Engineering, State of the Art and Current Trends
| Hi-index | 0.01 |
Analytic performance models are often used for predicting the performance of computing systems. Existing models accept single valued parameters as input and produce single valued performance measures as outputs. This research proposes to associate intervals or ranges of values with performance measures and key system parameters. Such an approach is appropriate when exact parameter values are unknown but approximate ranges for parameters may be estimated. Conventional arithmetic cannot handle intervals and interval arithmetic-based techniques are required. The paper reports on the feasibility of application of interval arithmetic in the solution of existing well-known models of computing systems. One of the problems with using interval arithmetic is the potential loosening in the interval for the model output: the computed interval may be wider than the actual interval. A computational method based on the notion of interval splitting is introduced in this paper for controlling this problem. The technique is found to be effective in the context of a number of models.