Publication Type : Conference Proceedings
Publisher : Companion Publication for ACM/SPEC on International Conference on Performance Engineering, ACM,
Source : Companion Publication for ACM/SPEC on International Conference on Performance Engineering, ACM, New York, NY, USA, p.45-50 (2016)
Url : https://dl.acm.org/doi/10.1145/2859889.2859898
ISBN : 9781450341479
Keywords : benchmarks, Cross-platform, performance, prediction
Campus : Amritapuri
School : Department of Computer Science and Engineering, School of Engineering
Center : Amrita Innovation & Research
Department : Computer Science
Verified : Yes
Year : 2016
Abstract : Predicting performance of multi-tier enterprise applications for a target platform is of significant importance to IT industries especially when target environment is unavailable for deployment. Performance modeling techniques depend on accurate estimation of resource demands for a specific application. This paper proposes a methodology for deriving Performance Mimicking Benchmarks (PMBs) that can predict resource demand of application server of multi-tier on-line transaction processing applications on a target environment. PMBs do not require the actual application to be deployed on the target itself. These benchmarks invoke similar method calls as the application at different layers in the technology stack that contribute significantly to CPU utilization. Further, they mimic all send and receive interactions with external servers (e.g., database server) and web clients. Ability of PMBs for service demand prediction is validated with a number of sample multi-tier applications including SPECjEnterprise2010 on disparate hardware configurations. These service demands when used in a modified version of Mean Value Analysis algorithm, can predict throughput and response time with accuracy close to 90%.
Cite this Research Publication : Subhasri Duttagupta, Kumar, M., and Apte, V., “Performance Mimicking Benchmarks for Multi-tier Applications”, Companion Publication for ACM/SPEC on International Conference on Performance Engineering. ACM, New York, NY, USA, pp. 45-50 , 2016.