Experimental evaluation and comparison of techniques, algorithms or complete systems is a crucial requirement to assess the practical impact of research results. The quality of published experimental results is usually limited due to several reasons such as: limited time, unavailability of standard benchmarks or shortage of computing resources. Moreover, achieving an independent, consistent, complete and insightful assessment for different alternatives in the same domain is a time and resource consuming task. We demonstrate Liquid Benchmark as a cloud-based service that provides collaborative platforms to simplify the task of peer researchers in performing high quality experimental evaluations and guarantee a transparent scientific crediting process. The service allows building repositories of competing research implementations, sharing testing computing platforms, collaboratively building the specifications of standard benchmarks and allowing end-users to easily create and run testing experiments and share their results
Liquid benchmarks: benchmarking-as-a-service
Casati, Fabio
2011-01-01
Abstract
Experimental evaluation and comparison of techniques, algorithms or complete systems is a crucial requirement to assess the practical impact of research results. The quality of published experimental results is usually limited due to several reasons such as: limited time, unavailability of standard benchmarks or shortage of computing resources. Moreover, achieving an independent, consistent, complete and insightful assessment for different alternatives in the same domain is a time and resource consuming task. We demonstrate Liquid Benchmark as a cloud-based service that provides collaborative platforms to simplify the task of peer researchers in performing high quality experimental evaluations and guarantee a transparent scientific crediting process. The service allows building repositories of competing research implementations, sharing testing computing platforms, collaboratively building the specifications of standard benchmarks and allowing end-users to easily create and run testing experiments and share their resultsI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione