Browse Prior Art Database

Process and Method for IT Energy Optimization Disclosure Number: IPCOM000015664D
Original Publication Date: 2002-Feb-21
Included in the Prior Art Database: 2003-Jun-20
Document File: 4 page(s) / 59K

Publishing Venue



Problem Summary

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 38% of the total text.

Page 1 of 4

Process and Method for IT Energy Optimization

Problem Summary

     Energy consumption and rising energy costs are an increasingly important concern for the profitability and stability of many companies. One important consideration that many companies overlook when planning their IT infrastructure is the regional fluctuations in energy prices. These fluctuations may be due to different energy providers, regional rates, geographical location and even the local climate. By allowing for regional differences in energy costs, IT infrastructures can make significant cost savings. The problem addressed in this disclosure is how to optimize energy cost across an IT infrastructure by leveraging the geographical location, regional energy prices, regional climates and technology characteristics.

Solution Overview

     This disclosure describes a method of mitigating energy cost by a method of workload management across a company wide clustered computing environment. Many companies employ clusters to run compute intensive, long running jobs and typically have many clusters scattered across the corporate IT infrastructure. This disclosure discusses a method of submitting jobs to clusters in an efficient way to optimize energy cost savings. This method would work in a corporate IT infrastructure for example, as well as a state university, or national labratory setting -- wherever cooperative clusters are scattered regionally.

     At a high level, the system would work as follows. During summer months in the north, much of the workload would be transferred to the southern hemisphere where the heat generated would be used to help warm the facility and reduce its heating consumption. During winter months in the north, work would remain north and be transferred from the south when the north has excess capacity. This system works because the more work being conducted, the more compute energy is required, which in turn translates into more online systems. It also works because the more technologies like CMOS are used, the more power they consume. The same principle would hold true in a state university environment. In New York state, for example, since energy costs are higher in New York City and Long Island, excess capacity in upstate facilites would be employed over running the same job downstate, where energy costs are higher.

     A novel embodiment to achieve the desired results is now shown. This system employs a two-level hierarchical job scheduler method. The top level scheduler (cluster scheduler) dispatches jobs to server farm clusters located in disparate geographies based on its unique algorithm. The second level scheduler (site scheduler) dispatches jobs within a single cluster at a site. The cluster scheduler enlists an entire cluster or a porition of a cluster for work, and assigns jobs to each cluster. The criteria by which the cluster scheduler determines which cluster is a combination of cluster availability, availability of suitable resources within t...