Performance and Resource Consumption Analysis of Elastic Systems on Public Clouds
Thiago Felipe da Silva Pinheiro, Francisco Airton Silva, Iure Fé, Danilo Mendonça Oliveira, Paulo Maciel · 2019
Service providers may build elastic systems on public clouds. The public cloud may offer economies of scale, but there are some considerations to take into account. Infrastructure-as-a-Service (IaaS) providers charge their customers by the use of virtual machines (VMs), and wrong deployment decisions may lead to financial losses. This paper proposes an approach for estimating systems' performance, use of VM instances and its related costs on a public cloud. This work proposes a Stochastic Petri Net (SPN)-based formal modeling strategy to represent elastic systems deployed on a public cloud and a cost model to predict the use of VM instances. The approach enables designers to plan and tune elastic architectures based on Mean Response Time (MRT) estimation. Using our strategy it is possible to estimate the impact of each deployment configuration on the evaluated metrics. Our modeling strategy considers reactive scaling policies. The model represents a remote infrastructure for supporting external users requests. By combining different instance types, simultaneous jobs per VM instance, stepsizes and scaling thresholds, it is possible to offer different response times for each deployment configuration. One case study was performed to evaluate the approach. Our approach has proven to be feasible and it highlights the most effective scenarios to minimize MRT and reduce costs.