Self-Optimising Decentralised Service Placement in Heterogeneous Cloud Federation
Emanuele Carlini, Massimo Coppola, Patrizio Dazzi, Matteo Mordacchini, Andrea Passarella · 2016
Clouds have been relevant business drivers and computational backends for a wide range of applications, including IoT, e-health, data analytics. To match the complex needs of such comprehensive set of different kinds of applications, in recent times there is an emerging need for new paradigms and forms of Clouds, organised according to a federated, heterogeneous and distributed structure. To exploit heterogeneity and localisation, in order to enhance the overall performances, ensure energy efficiency, reduce costs for resource providers and in the meantime enhance the user experience, proper service placement solutions are required. However, conducting efficient deployments in such a scenario is complex due to the dynamic nature of applications, resources, users. As a consequence, there the a need for scalable, distributed, adaptive, context-aware solutions characterised by high-efficiency and reduced overhead. We propose a highly distributed, self-adaptive solution aimed at optimising the overall deployment of cloud services by means of point-to-point interactions occurring among clouds and cloudlets belonging to the same federation. The contribution of this paper is the definition of a service exchange mechanism, its Markov-chain based modelling and thorough experimental evaluation.