Search

Efficient resource management is crucial for balancing performance and energy consumption in large-scale data centres. In the case of additional requirements such as guaranteed resources and low communication latency, it is of great importance to implement not only an efficient initial placement algorithm, but also maximise consolidation by migration techniques, making sure that network performance is not sacrificed. In this paper, we introduce a hierarchical approach to migrations based on a combination of efficient packing algorithms and network communities. Results analysis shows the benefits of using a two-level approach where the combination of localised consolidation and network awareness improves both performance and energy efficiency, while maintaining low network hop distance.

The ongoing trends in cloud computing demonstrate increasing need for efficient, yet economical data centers. Thus, recently the research community has focused its efforts on frameworks for optimised usage of the available resources that will result with energy-efficient and highly effective data centers. Toward this goal, in this paper we present a community-based framework for virtual machine placement inside a cloud data center. The framework is based on the complex network structural property of grouping tightly coupled nodes, and a matching process that maps virtual to physical communities while employing different optimisation functions on different hierarchy levels. The presented simulation results of the framework application reflect its high usage potential achieved by improvement in communication efficiency and reduced power consumption compared to the traditional heuristics.