ABSTRACT
In general, small and medium-scale enterprises (SMEs) face problems of un-predictable IT service demand and infrastructure cost. Thus, the enterprises 344strive towards an IT delivery model which is both dynamic and flexible, and able to be easily aligned with their constantly changing business needs. In this context, Cloud computing has emerged as a new approach allowing anyone to quickly provision a large IT infrastructure that can be completely customized to the user’s needs on a pay-per-use basis. This paradigm opens new perspectives on the way in which enterprises’ IT needs are managed. Thus, a growing number of enterprises are outsourcing a significant percentage of their infrastructure to Clouds.
However, from the SMEs perspective, there is still a barrier to Cloud adoption, being the need of integrating current internal infrastructure with Clouds. They need strategies for growing IT infrastructure from inside and selectively migrating IT services to external Clouds in a way that enterprizes benefit from both Cloud infrastructure’s flexibility and agility as well as lower costs.
In this chapter, we present how to profitably use Cloud computing technology by using Aneka, which is a middleware platform for deploying and managing the executions of applications on different Clouds. This chapter also presents public resource provisioning policies for dynamically extending the enterprise IT infrastructure to evaluate the benefit of using public Cloud services. This technique of extending capabilities of enterprise resources by leasing public Cloud capabilities is also known as Cloud bursting. The policies rely on a dynamic pool of external resources hired from commercial IaaS providers in order to meet peak demand requirements. To save on hiring costs, hired resources are released when they are no longer required. The described policies vary in the threshold used to decide when to hire and when to release; the threshold metrics investigated are queue length and queue time as well as a combination of these. Results demonstrate that simple thresholds can provide an improvement to the peak queue times, hence keeping the system performance acceptable during peaks in demand.
