Mapping Applications on Computing Clouds and Service-Oriented Systems

Computing Clouds have been promoted in recent years as architectures which provide scalable resources to satisfy user demand for particular services. In such architectures, there is a provider, who owns the resources, and there are consumers who need to use services available from these resources. Upon a user’s request to use a service, the provider has to examine whether the request can be satisfied and, if yes, to offer an agreement, commonly known as a service-level agreement.

In such an environment, users may have conflicting requests. For example, a user may prefer to run an application as quickly as possible, another user may prefer to pay as little as possible and so on. From the provider’s point of view there is an issue of optimizing the allocation of different applications onto resources both in order to satisfy service-level agreements and maximize profit.

Earlier work has considered the use of service-level agreements as an alternative to traditional high-performance parallel computing [1]. Other work has considered issues related to mapping workflow applications. The objective of this project is to expand this research by considering more dynamic, service-oriented environments, where multiple types of services may be running. The key challenge will be to model and investigate techniques that can allow these environments to cope efficiently with conflicting constraints such as response time, user’s budget, energy consumption and more.