SUPPORTING APPLICATION QUALITY OF SERVICE IN SHARED RESOURCE POOLS

摘要

Many enterprises are beginning to exploit large shared resource pools in data center environments to lower their infrastructure and management costs. These environments may have tens, hundreds, or even thousands of server resources. Capacity management for resource pools decides how many resources are needed to support a given set of application workloads, which applications must be assigned to each resource, and per-application scheduling parameters to ensure appropriate sharing and isolation for the applications. Capacity management is a challenging task for shared environments that currently requires significant manual effort and tends to overprovision resources. Here, we describe our approach to automate the steps of a capacity self-management system that best matches resource supply with demand.