Network-aware placement of virtual machine ensembles using effective bandwidth estimation

Modern datacenters rely heavily on virtualization technologies to offer customized computing and network resources on demand to a large number of tenant applications. However, efficiency in resource utilization delivered by virtualization technologies that exploit statistical multiplexing of resources across applications means that predictability in performance remains a challenge. Allocation of network bandwidth is particularly difficult, given the variability of traffic flows between the components of multi-tier applications. Static bandwidth allocation based on peak traffic rates ensures SLA compliance at the cost of significant overprovisioning, while allocation based on mean traffic rates ensures efficient usage of bandwidth at the cost of QoS violations. We describe MAPLE, a network-aware VM ensemble placement scheme that uses empirical estimations of the effective bandwidth required between servers to ensure that QoS violations are within targets specified in the SLA for the tenant application. Experimental results obtained using traffic traces collected from an emulated datacenter show that, in contrast to the Oktopus network-aware VM placement system, MAPLE is able to allocate computing and network resources in a manner that balances efficiency of resource utilization with performance predictability.