Implementation of a Fast Vector Packing Algorithm and its Application for Server Consolidation

With increasing adoption of SOA and Cloud Computing technologies where IT including infrastructure, platforms and applications are delivered as services, there is increasing use of a shared resource model where computing and IT resources are shared across multiple applications, so accordingly there is increasing need for solutions that optimize the resource allocation. Power, cooling and real estate are significant costs in operating a cloud computing platform so there is need for solutions that optimize the resources consumed in order to reduce these costs. The challenge in these is in consolidating workloads to minimal number of servers while taking into consideration the resource needs across multiple dimensions like compute, storage, IO, networking bandwidth, etc which keeps changing continuously. This is considered to be a NP hard problem for which there are several solutions based on traditional bin packing algorithms. These solutions have limitations in arriving at the optimal solution in short enough time to be able to react to changing workloads. We describe an algorithm that enables arriving at an optimal workload consolidation solution with desired accuracy by trading off the accuracy with the processing required to arrive at the optimal solution while taking into consideration multiple resource usage dimensions like CPU usage, IO usage, network bandwidth usage etc simultaneously to arrive at the optimization.