From Simulation to Experiment: A Case Study on Multiprocessor Task Scheduling
Year: 2011
Type: article
Abstract: (e.g., a few minutes of simulation on a single computer could suffice for simulating an application that runs for several hours on hundreds of cluster nodes). Consequently, the vast majority of published simulation results in the parallel application scheduling literature are obtained with analytical simulation models. In particular, simulation tools used routinely in the community use analytical, some of which have been validated against operational models [2]. The key question we investigate in this work is whether relying on analytical simulation models leads to scientifically valid conclusions. This is a somewhat disturbing question to ask given the sheer number of published results that rely on such models. It is also a broad question that does not have a single answer since the answer depends on the applications and platforms at hand. Consequently, we focus on a relevant case
Authors: Sascha Hunold, Henri Casanova, Frédéric Suter
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2978995165