Penalty Model for SLA Architecture in Cloud

Abstract

Cloud computing is emerging as a commercial infrastructure for providing the computing facilities on par with Supercomputers. It has the capability to scale up to „n‟ number of processes or threads as required by the application developers. The cloud concept has motivated the application developers to host their applications as Cloud service with the service providers such as Amazon, GoGrid, Mosso, Google Cloud, and so forth. To host such applications, in general, they sign up a Service Level Agreement (SLA) with the providers to ensure high availability (99.99%), security, and reduce managerial issues. Similarly, the cloud users who utilize the cloud services sign up the SLA as a part of a service contract where the level of service is formally defined. The service contract could explicitly specify the number of cores available for a certain time, the CPU speed, the network bandwidth, the cache behaviour, the percentage of parallelization, and so forth. In most of the cases, however, the performance is not met as per the agreement signed between the service providers and the service consumers. The challenge is more crucial when the number of consumers increases or when the demand for resources increases. This paper proposes a penalty-based SLA architecture for cloud environments. In case of violation by the service provider this model imposes penalty measuring the degree of violation. In addition, the paper highlights the merits and demerits of the existing architectures in SLA-based clouds.