Dynamic Scheduling based on Particle Swarm Optimization for Cloud-based Scientific Experiments

Abstract

Parameter Sweep Experiments (PSEs) allow scientists to perform simulations by running the same code with different input data, which results in many CPU-intensive jobs, and hence par- allel computing environments must be used. Within these, Infrastructure as a Service (IaaS) Clouds offer custom Virtual Machines (VM) that are launched in appropriate hosts available in a Cloud to handle such jobs. Then, correctly scheduling Cloud hosts is very important and thus efficient scheduling strategies to appropriately allocate VMs to physical resources must be devel- oped. Scheduling is however challenging due to its inherent NP-completeness. We describe and evaluate a Cloud scheduler based on Particle Swarm Optimization (PSO). The main performance metrics to study are the number of Cloud users that the scheduler is able to successfully serve, and the total number of created VMs, in online (non-batch) scheduling scenarios. Besides, the number of intra-Cloud network messages sent are evaluated. Simulated experiments performed using CloudSim and job data from real scientific problems show that our scheduler achieves better performance than schedulers based on Random assignment and Genetic Algorithms. We also study the performance when supplying or not job information to the schedulers, namely a qualitative indication of job length.