A bio-inspired scheduler for minimizing makespan and flowtime of computational mechanics applications on federated clouds

Abstract

Computational Mechanics (CM) concerns the use of computational methods to study phenomena under the principles of mechanics. A representative CM application is parameter sweep experiments (PSEs), which involves the execution of many CPU-intensive jobs and thus computing environments such as Clouds must be used. We focus on federated Clouds, where PSEs are processed via virtual machines (VM) that are lauched in hosts belonging to different datacenters, minimizing both the makespan and flowtime. Scheduling is performed at three levels: a) broker, where datacenters are selected based on their network latencies via three policies, b) infrastructure, where two bio-inspired schedulers based on Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO) for VM-host mapping in a datacenter are implemented, and c)VM, where jobs are assigned into the preallocated VMs based on job priorities. Simulated experiments performed with job data from two real PSEs show that our scheduling approach allows for a more agile job handling while reducing PSE makespan and flowtime.