?
A Versatile Simulator for Complex Cluster Workloads
Resource allocation and scheduling algorithms play a crucial role in operation of modern computing clusters. The algorithms used and their settings can significantly impact cluster utilization, application performance, and user experience. This area of research requires the use of simulators to quickly test a hypothesis or conduct comparative testing of different algorithms. This paper presents ClusterSim, a flexible and fast simulator of computing cluster and workloads. In contrast to other existing simulators, it supports both detailed modeling of complex workloads and fine-grained resource allocation, which allows to apply it for various problem statements and scenarios. At the same time, the basic workload scenarios are also supported, in which high performance is achieved. The conducted experiments demonstrate the applicability of ClusterSim for running large-scale simulations, modeling different scheduling approaches, and studying complex problem statements.