Greedy Performance Metrics for Grid Schedulers
The most common metrics to analyse the effectiveness of schedulers from the user’s perspective are response time, waiting time and slowdown. These metrics are used to understand how good are the scheduling algorithms, and their main objective is to measure how quick the jobs are processed. However, these basic metrics are not fair for schedulers of grids and heterogeneous clusters. In this paper we introduce new metrics specially adapted to heterogeneous systems. They can be considered as a generalisation of the previous ones. The benefits and drawbacks of these metrics are also studied from both, theoretical and practical points of view.
keywords: Grid computing, performance metrics
Publication: Congress
1624015015874
June 18, 2021
/research/publications/greedy-performance-metrics-for-grid-schedulers
The most common metrics to analyse the effectiveness of schedulers from the user’s perspective are response time, waiting time and slowdown. These metrics are used to understand how good are the scheduling algorithms, and their main objective is to measure how quick the jobs are processed. However, these basic metrics are not fair for schedulers of grids and heterogeneous clusters. In this paper we introduce new metrics specially adapted to heterogeneous systems. They can be considered as a generalisation of the previous ones. The benefits and drawbacks of these metrics are also studied from both, theoretical and practical points of view. - J. L. Albín, T. F. Pena, J. C. Cabaleiro and F. F. Rivera - 10.4203/ccp.90.23
publications_en