Joint Virtual Machine Placement and Path Selection in a Virtualized Datacenter Environment

Today's multi-tenant shared datacenter environments guarantee compute resources over a best-effort intra-datacenter network. Bursty traffic in such a network implies higher delays and losses for tenants. This may result in unpredictable network performance, potentially hurting the tenant's overall application performance. Motivated by these concerns, in this work we propose a Service Level Agreement (SLA) that along with compute capacity, explicitly guarantees intra-datacenter network bandwidth to the tenants in a virtualized, multi-tenant datacenter environment. We design a modified greedy approach that uses virtual machine placement and path selection to solve the joint resource allocation problem. In recent years, several algorithms have been proposed for resource allocation problems to meet a specific SLA. However they do not provide any practical recommendations for the realization of these solutions from a datacenter designer's standpoint. Another main contribution of our work is the design of a rigorous simulation setup to derive a definitive set of guidelines for the datacenter designer implementing the proposed SLA. The guideline answers questions of key practical importance to datacenter designers and enables derivation of effective estimates about the switch fabric and server capacity to be bought and installed, as a function of tenant demands.