The Effect of Workload Computational Demand Variability on the Performance of a SaaS Cloud with a Multi-tier SLA

• Published in: 2017 IEEE 5th International Conference on Future Internet of Things and Cloud (FiCloud) pp. 10 - 17
• Main Authors: Stavrinides, Georgios L., Karatza, Helen D.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2017
• Subjects: bags-of-tasks; Cloud computing; Computational modeling; multi-tier SLA; performance evaluation; Processor scheduling; Quality of service; SaaS clouds; scheduling; Software as a service; Time factors; Workload computational demand variability
• DOI: 10.1109/FiCloud.2017.26

In the highly competitive business environment of Software as a Service (SaaS) clouds, Quality of Service (QoS) and fair pricing are of paramount importance for differentiating between similar cloud providers. In such platforms, the workload computational demand variability may have a significant impact on the system performance and thus on the provider's Service Level Agreement (SLA) commitments. Towards this direction, in this paper we consider a SaaS cloud with a multi-tier SLA that focuses on the fair billing of the end-users, according to the provided level of QoS. The workload consists of bag-of-tasks jobs, which are scheduled on the underlying virtualized host environment. The jobs have soft deadlines and different levels of variability in their computational demands. The effect of the workload computational demand variability on the system performance, under the employed multi-tier SLA, is investigated via simulation. The experimental evaluation provides useful insights into the impact of computational demand variability on the performance of the SaaS cloud under study.