Automatic provisioning of intercloud resources driven by nonfunctional requirements of applications

Cloud computing enables a major paradigm shift in the way that computing resources are acquired. Without any hardware acquisition, system administrators are able to obtain computing power to deploy their services within minutes using cloud computing. It also makes it possible to pay only for resources consumed with no minimum contract and upfront costs. However, deploying applications in the cloud is still a complex task for system administrators. They are expected to estimate resources required by their applications, which may be difficult because they frequently do not know exactly how many resources are actually necessary. Furthermore, administrators have to select the best cloud service amongst various providers and different types of services, and acquire them for applications to deliver the expected performance. In this chapter, we proposed an architecture supporting system administrators in the arduous task of deploying applications on the clouds in three ways. Firstly, it translates nonfunctional requirements from administrators into cloud resource parameters. Secondly, it selects the most convenient provider among different candidates, which satisfies every requirement. Finally, the actual VMs are allocated automatically from the selected provider. The proposed architecture is also verified through evaluation and validation. Each component is validated in performance and scalability for various sets of nonfunctional requirements. Also, we show that the number of VMs with adequate resources is actually allocated from the selected cloud provider at the end of the process. We recommend a few areas for further work. The proposed architecture can be applied to measure the performance of various techniques in each module. Several resource estimation techniques can be applied to the estimator in the architecture, which can lead to the most accurate methodology to estimate resources in the clouds being found. Similarly, different approaches to select the best provider can be used for the selector module. In addition, the cloud information database can be improved by applying dynamic updates; thus, it will keep the consistency between the system and the providers, and will provide more accurate selection by including real‐time metrics measured by monitoring services. Finally, dynamic resource provisioning can be applied to the system, which can dynamically perform the whole provisioning process depending on the real‐time workload measured from the running application