This Redpaper describes a proof of concept (PoC) project that was conducted in the Poughkeepsie, NY, Design Center for On Demand business. The PoC was developed for XYZ Automotive Company (XYZ). XYZ representatives worked with IBM IT Specialists and Architects to define critical success factors, use cases, and requirements for the project.
XYZ has an environment where engineers use grids for their Computer Aided Engineering work. They have many heterogeneous local grid schedulers. The engineers schedule their work to two islands of grids: a Numerically Intensive Compute center, and a Non-Dedicated Grid. Engineers must specifically select the domain or node for the job to run, and are able to select number of CPUs, memory, and disk space. However, there are no policies or rules to block the engineer’s requests. This leads to underutilization of the XYZ environment. This type of environment is prevalent in sectors where grid computing is at a mature level. Departmental grid and high performance computing clusters have been put in place to solve various compute-intensive problems. Grids made available from mergers and varying lines of business have resulted in a heterogeneous grid environment, where silos of grid compute power are clearly not being optimized across the enterprise. A metascheduler can solve these problems. A metascheduler provides a consistent interface to multiple heterogeneous grids. In this paper we discuss the requirements for a metascheduler for a customer in the automotive industry.
Table of contents