CART, a Decision SLA Model for SaaS Providers to Keep QoS Regarding Availability and Performance

Cloud systems are becoming a powerful tool for business. The evidence of the advantages of offering infrastructure as a service (IaaS), platform as a service (PaaS), or software as a service (SaaS) is overwhelming. Therefore, for SaaS providers, it is essential to know the virtual resources required to optimize the service offered and to keep the quality of service (QoS) at the desired levels. This paper presents an analytic model cloud availability and response time (CART) for obtaining the best tradeoff between performance, cost, and availability in a cloud system aimed at providing software as a service. The model aims to guarantee the predetermined availability and response time (RT) agreed with customers in a service-level agreement (SLA) contract while minimizing the cost of the system. A client-transparent error recovery strategy has been considered along with a Poisson traffic model. A sensitivity analysis of the simulated and real workloads is presented to study how a specific SLA influences the cloud service performance regarding the guaranteed RT and availability. The results corroborate the goodness of the model proposed, adjusting the real system accurately. Furthermore, the obtained results provide useful guidelines for cloud or Web-server designers.