Executive Summary

From The Business of Energy in U.S. Academic Research Computing
Jump to: navigation, search

This project was completed in order to explore two major areas of inquiry. First was the applicability of business concepts to the industry of academic supercomputing. Second was an exploration of energy as it applies within the industry. This exploration relied heavily on prior work by CASC members and others who have pursued ongoing interests in sustainability and resiliency for high performance computing (HPC) in higher education.

On April 25, 2014 I presented a project summary and highlights of the accompanying spreadsheet to the spring meeting of the Coalition for Academic Scientific Computing (CASC). The major deliverables of the project consisted of an online monograph and an accompanying spreadsheet. Both are freely available online at http://hpcbiz.readingroo.ms.

In the area of business concepts, the monograph includes an industry analysis, discussion of key productivity indicators (KPIs) and stakeholders, discussion of employee development, and other topics. The premise of the discussion is that HPC managers are frequently required to justify the return on investment created by their centers. Grounding in the language and concepts of modern business is useful for making the case for a strong ROI. The spreadsheet provides a template for tracking ROI for a center within a particular institutional environment, along with weighted aggregations of measures for presentation to management.

In the area of energy for high performance computing, the analysis starts with assessing energy use by high performance computing systems, and the fact that digital computation, by its nature, produces heat that must be removed. The spreadsheet also includes a calculator for total lifetime energy costs for operation. Additional calculators estimate the embedded energy (the energy required to produce the system) and the mix of raw materials in the system. These energy values are input to another calculator, which estimates carbon dioxide output due to generating the electricity used throughout the lifetime of the system.

Major outcomes emerging from the project include:

  1. The value of maintaining return on investment (ROI) estimates for an HPC center, where calculations are based on outcomes deemed pertinent for the host institution. Such outcomes are often non-financial. However, they may be measured, tracked over time, and compared to other organizations at the institution.
  2. Applying business concepts for staff development within a center, even without explicit direction from the host institution, can yield improvements to employee career path options and employee retention.
  3. It is straightforward to quantify the approximate ongoing costs of operating a supercomputer, such that those costs may be compared for different scenarios of systems upgrades or acquisitions. Because costs per unit of computation (e.g., dollars per gigaFLOP) decrease over time as technologies improve, a new computer's costs are offset by decreased energy use.
  4. A decision to upgrade components (i.e., system boards and interconnect), versus replacing an entire system (including the metal chassis, cabinets, and other physical superstructure) can yield significant savings in total embedded energy.

I wish to thank the CASC project sponsors, Chair David Lifka, Sue Fratkin, and Nicholas Berente, for guidance and input during this effort. I also wish to thank the many members of CASC who offered input to me, or via participation in two NSF sustainability symposia. I thank my instructors and student colleagues at the Bainbridge Graduate Institute of Pinchot University. Finally, I thank my wife Ilana for ongoing encouragement and support.