EFFICIENT PLACEMENT OF VIRTUALIZED INSTANCES WITH INTERGROUP AFFINITY AND ANTI-AFFINITY

Abstract

A data processing environment has a set of consumer pods and a set of provider nodes, and the set of consumers includes a first subset of a first group of consumers and a second subset of a second group of consumers. A pricing function in a solver component in the environment is configured with an intergroup affinity between the first subset and the second subset of consumers such that a member of the first subset of consumers gets placed on a selected provider from the set of providers when the selected provider already hosts a member of the second subset of consumers. The solver uses the intergroup affinity to cause a first placement of a first consumer from the first subset at a first provider from the set of providers, when the first provider already hosts a first consumer from the second subset of consumers.

CPC Classifications