Elastic service definition in computational clouds

Cloud computing is rapidly changing the landscape of traditional IT service provisioning. It presents service providers with the potential to significantly reduce initial capital investment into hardware by removing costs associated with the deployment and management of hardware resources and enabling them to lease infrastructure resources on demand from a virtually unlimited pool. This introduces a great degree of flexibility, permitting providers to pay only for the actual resources used per unit time on a “pay as you go” basis and enabling them to optimise their IT investment whilst improving the overall availability and scale of their services. In this manner, the need for overprovisioning of services to meet potential peaks in demand can be considerably reduced in favour of driving resource allocations dynamically according to the overall application workload. This however requires means of defining rules by which the service should scale and mechanisms must be provided by the infrastructure to monitor this state and enforce the rules accordingly. We discuss in this paper an elastic service definition language and service management that are developed in the context of the RESERVOIR project to facilitate dynamic service provisioning in clouds. This language builds on the Open Virtualisation Format, a DMTF standard for the packaging and deployment of virtual services, and introduces new abstractions to support service elasticity. In addition, we detail the functional requirements that the the cloud infrastructure must meet to handle these abstractions.