Challenges in the Management of Federated Heterogeneous Scientific Clouds

Abstract

Cloud computing has evolved towards a model of computing that is already used widely, and often is the choice for new applications and services that are to be built. The trend is not only observable in commercial applications but also in academic ones. Since the beginnings of cloud computing deployment models of federated and hybrid clouds have always been part of the discussions. However, in practice federated clouds are hard to achieve since cloud stacks tend to be heterogeneous and lock their customers in. In this work we discuss an architecture that adds a layer of abstraction on top of the existing architecture of infrastructure clouds and makes implementation details of different cloud stacks transparent to the user. This way we can mitigate some problems that are introduced with distributed federated heterogeneous clouds. Further we introduce a process that allows for wide-area virtual machine migration on top of this architecture and show why this approach is feasible in practical applications.