Analytical evaluation of resource allocation algorithms and process migration methods in virtualized systems

In this paper, analytical models, based on the stochastic reward nets (SRNs), are proposed to analyze the impact of resource allocation algorithms and process migration methods on the power consumption and performance of virtualized systems. In the proposed models, each computer offers a certain capacity of computational, data, and communication resources and resource requesters, namely processes, are categorized into three main categories: compute-intensive, data-intensive and communication-intensive, according to their required resources. Since the processing rate of a computer reduces when the number of the processes running on the computer increases, we apply the migration methods to keep the performance of the system at a high level and reduce the power consumption as much as it is possible. The proposed SRNs appropriately model the migration of processes among computers by applying two types of migration methods: power-aware and performance-aware. Furthermore, by applying the proposed models, different resource allocation algorithms, e.g. First-fit, Best-fit, Worst-fit, and Random, can be compared in terms of the power consumption and performance. The numerical results, cross-validated with the CloudSim framework, show that the proposed models can be appropriately used to analyze different resource allocation algorithms and migration methods, and thereby, help system providers to make treatment decisions with confidence. (C) 2019 Elsevier Inc. All rights reserved.