Strategic Provision of Cloud Computing Services

Research Issues and Challenges

This section presents several important open issues and research challenges as well as research directions for successful service provisioning deployment and implementation.

Service availability becomes more important in a dynamic environment. Applications require intensive interaction between the end-user and the cloud service. Hence, service disruption, network congestion, poor signal, and node failure are highly undesirable in service provisioning. When a node moves frequently within the network or changes its point of attachment, many mobile cloud applications demand optimal service through the most suitable node. Again, the scalability of services is a challenging aspect of distributed application processing in cloud computing. Remote application processing is deficient in the centralised management of the distributed platform. A challenging issue in local distributed application frameworks (APFs) is the unavailability of centralised resources. When a remote service provider is unavailable, remote services become inaccessible, which hinders the objectives of availability of services in a distributed computing paradigm. In addition, the QoS requirements are also evolving with the evolution of the cloud, and therefore service provisioning requires highly reliable good service quality. Similar services and functionalities are provided by the different CSPs, which makes it difficult for customers to select the best and most appropriate service. Optimal provider selection based on predetermined quality of service (QoS) requirements becomes vital. Moreover, the users' past experiences are exploited in a ranking-based approach to accumulate and identify the preferences between pairs of services to obtain a service ranking. Thus, this approach gets benefits from past user experience. However, user response does not always reflect true feedback. It is necessary to adjust and justify the feedback using statistical techniques to make it as error-free as possible. Furthermore, SLAs are legally binding for both parties, recognised as the terms and conditions of using the service. Because of the fast-growing number of promising service offers and the lack of a standard specification of services, manual service selection is an expensive task, preventing the successful implementation of ubiquitous computing on demand. Therefore, automatic methods for matching SLAs are essential. Finally, several mathematical and statistical methods and model-based issues are proposed and validated. Service provisioning requires a precise and efficient artificial intelligence learning mechanism, and in the planning stage the number of variants, resource constraints, bandwidth limitations, and properties should of course be considered with respect to run time, hardware, and software.