Stochastic Models for Planning VLE Moodle Environments based on Containers and Virtual Machines

Abstract

Moodle Virtual Learning Environments (VLEs) represent tools of a pedagogical dimension where the teacher uses various resources to stimulate student learning. Content presented in hypertext, audio or video formats can be adopted as a means to facilitate the learning. These platforms tend to produce high processing rates on servers, large volumes of data on the network and, consequently, degrade performance, increase energy consumption and costs. However, to provide efficient sharing of computing resources and at the same time minimize financial costs, these VLE platforms typically run on virtualized infrastructures such as Virtual Machines (VM) or containers, which have advantages and disadvantages. Stochastic models, such as stochastic Petri nets (SPNs), can be used in the modeling and evaluation of such environments. Therefore, this work aims to use analytical modeling through SPNs to assess the performance, energy consumption and cost of environments based on containers and VMs. Metrics such as throughput, response time, energy consumption and cost are collected and analyzed. The results revealed that, for example, a cluster with 10 replicas, occupied at their maximum capacity, can generate a 46.54% reduction in energy consumption if containers are used. Additionally, we validate the accuracy of the analytical models by comparing their results with the results obtained in a real infrastructure.