Cost and performance analysis of MPI-based SaaS on the private cloud infrastructure

Abstract

The paper presents the cost and performance analysis of parallel MPI-based software as a service (SaaS) deployed on the OpenStack cloud infrastructure. The parallel SaaS was developed by using C++ programming language and MPI library for the scientific discrete element method (DEM) computations of granular flows. The performance measured on KVM-based virtual machines was slightly higher than that on Docker containers of the OpenStack cloud. Round up and proportional pricing schemes were examined and compared from the user’s perspective. The difference in cost computed by using alternative pricing schemes varied from 0.6% to 15.4%. However, this difference can be reduced to 1.0%, increasing execution time of considered tasks. The investigation of a trade-off between the execution time and cost was performed by using Pareto front analysis and a linear scalarization method. Bi-objective decision making revealed the preferable configurations of virtual machines specific to memory bound DEM computations, exploiting higher bandwidth.