Dynamic application deployment in federations of clouds and edge resources using a multiobjective optimization AI algorithm
Ram Govinda Aryal, Jörn Altmann · 2018
Cloud federation brings, besides various opportunities, challenges for resource allocation. Optimized allocation of resources to applications is one of these challenges. The challenge is further enhanced by the heterogeneity of provider resources, which can be edge resources or cloud resources, and by the perception of the relative importance of optimization objectives. What is required by a federation broker, in such a context, is an efficient decision-making algorithm for virtual machine (VM) placement. Such an algorithm should first identify eligible resources and, then, select optimal combinations of resources within the federated cloud, considering multiple optimization objectives that are derived from the application requirements. The relative weights of which should be tuned to meet individual application needs. Furthermore, it should be able to work dynamically throughout the application lifecycle rather than only during the initial deployment to address changes in application and user behaviors. This paper proposes such an algorithm for the BASMATI cloud federation architecture. The results of multiple runs of our simulation demonstrate that the algorithm, which is a genetic algorithm, is efficient, can provide optimal solutions for VM placement decision making, and can be tuned to address specific application needs.