Output details
11 - Computer Science and Informatics
University of St Andrews
Decentralized and optimal control of shared resource pools
<02>The capstone journal paper of a series addressing a crucial question in cloud computing: how to allocate shared resources optimally under variable workloads? It presents a ground-breaking fully-decentralised solution to this problem, suitable for application at data-centre scales. A proven-converging mathematical optimisation model is implemented as a multi-agent system, using efficient epidemic spreading to underpin efficient distributed calculations. The system is evaluated against synthetic data centre workloads and is shown to converge to globally optimal resource allocation even in the presence of failing nodes, and with best-effort service of each individual processes' resource requirements.