Stochastic Petri Net Models for Availability and Performance Evaluation of Nextcloud Service hosted in Apache Cloudstack

• Wenderson Leonardo UFRPE
• Thiago Bezerra IFPE
• Gustavo Callou UFRPE

Abstract

Technology is increasingly becoming an integral part of human daily life, and as a direct result, there is a growing desire to address everyday needs through applications. These applications require high availability levels and satisfactory performance responses. In this context, cloud computing, with its elasticity and on-demand resource access, represents an interesting solution to support such applications. Besides that, given that companies are inclined to migrate to private cloud environments, it is necessary to understand the optimal manner for allocating their cloud instances. This study proposes models in stochastic Petri nets to evaluate the Nextcloud service hosted in the private cloud Apache Cloudstack in terms of metrics such as availability, throughput, and response time. Nextcloud is a suite similar to Dropbox, Office 365, or Google Drive that offers client-server software for creating and using file services. A case study is proposed to analyze the performance and assess the optimal method for distributing virtual machines (VMs) that host such a service. The achieved results show that increasing the number of instances hosting the service leads to improvements in both availability and performance metrics. For instance, adding one virtual machine (VM) containing the service yields an improvement of up to 4.65% in availability, 1.2% in throughput, and 1.05% in response time. Moreover, the addition of one VM with a load balancer results in an improvement of up to 5.1% in availability, 2.03% in throughput, and 1.94% in response time.