Provisioning Norm: An Asymmetric Quality Measure for SaaS Resource Allocation

Large scale, multi-dimensional resource provisioning for Software as a Service (SaaS) presents a significant challenge. Analytical calculation of the quality of a configuration is necessary for effectively assigning new services to servers and reorganizing assigned services. This work describes the Provisioning Norm which meets this critical need. The Provisioning Norm (an asymmetric norm) analytically calculates the quality of a configuration, the placement of services on server nodes. The Provisioning Norm partially orders all possible configurations for a set of nodes and services from best to worst by numerically biasing over-provisioned configurations relative to under-provisioned configurations. This work proves that the parameter to the Provisioning Norm function has a value which partitions the partially ordered configurations into over-provisioned configurations and under-provisioned configurations. The application of the Provisioning Norm in a testing environment demonstrates a correlation between the analytical quality and the empirical performance.