Graph-Based Cloud Service Placement

The emergent IT clouds as the future of datacenters enable considerable opportunities for the services creation, deployment, management and usability. Users all over the world, from individuals to businesses have been taking advantage of the new cloud services automation and scalability benefits. However, the services creation and business support are still dominated by intensive manual labor. Offerings with similar infrastructure requirements and dependencies are mainly built from scratch as separated entities, making the service development inefficient and error prone. We propose a graph based solution for semi-automated service creation, which expresses the mapping between a business support system and an operations support system. We first identify and expose, at the leaf level of our graph, the meaningful IT operations in the form of basic services. Then, we extend our graph by representing existing services offerings in terms of these operation level service definitions as well as simpler services offerings. At service creation time, an offering manager can re-combine existing building blocks to define new services, besides implementing new blocks down to the operations support system. Our solution takes into consideration the constraints and costs of the service offering sub-components as far as their mapping down to datacenter resources for optimizing the service placement into data-centers. We present a study of the Desktop Service use case.