Title: The Power Of Machine Learning And Market Design For Cloud Computing Admission Control
Abstract: Cloud computing providers must handle customer workloads that wish to scale their use of resources such as virtual machines up and down over time. Currently, this is often done using simple threshold policies to reserve large parts of compute clusters, which leads to a low average utilization of the cluster. We present new, more sophisticated policies that can take learned or elicited probabilistic information about the future behaviors of arriving workloads into account. Based on these policies, we show the high potential machine learning and market design have for increasing efficiency.
