Skip to content

Modernizing Enterprise Data Centers for Fun and Profit

Jonathan Koomey Jonathan Koomey

Twenty first century data centers are the crown jewels of global business. No modern company can run without them, and they deliver business value vastly exceeding their costs. The big hyperscale computing companies (like Google, Microsoft, Amazon, and Facebook) are the best in the industry at extracting that business value, but for many enterprises whose primary business is not computing, the story is more complicated.

If you work in such a company, you know that data centers are often strikingly inefficient. While they may still be profitable, their performance still falls far short of what is possible. And by “far short” I don’t mean by 10 or 20 percent, I mean by a factor of ten or more.

money waste trash thinkstock cookelma

Source: Thinkstock / cookelma

A SHOCKING WASTE

The waste embodied in most enterprise data center operations should be shocking to anyone who cares about financial performance. In a recent study of 4000 enterprise servers, my colleagues and I found that about 30 percent of these servers were comatose (using electricity but delivering no useful computing), a result consistent with results from McKinsey and the Uptime Institute. McKinsey also found that typical utilization for servers in enterprises “rarely exceeds” 6 percent. Many enterprises don’t even know how many servers they have, and couldn’t tell you their average utilization if they tried.

Why do such inefficiencies persist? The reasons are not primarily technical, but revolve around people, institutions, and incentives, and start with fractured management chains that have separate budgets for IT and facilities departments. Poor measurement, misplaced incentives, and failure to apply best-in-class technologies to full advantage lead to most data centers underperforming.

What can management do? First, centralize data center operations, with one boss, one team, and one budget. Only then will the team focus on the whole system costs and benefits of any proposed changes.

Second, tie data center performance to business performance, mapping data center infrastructure costs onto business processes, and using metrics that show the business implications of data center choices. Every part of the business should be able to compare total IT costs to benefits at the project level. Most importantly, companies should calculate – or at least, estimate – total costs and revenues per computation.

Finally, companies must use the power of IT to transform IT. The most advanced companies apply measurement and engineering simulation to optimize data center operations, tracking server inventories, monitoring real-time conditions, and projecting the impacts of proposed new IT deployments.Standardizing on a few server designs instead of dozens or hundreds reduces deployment times from months to days, and moving smaller computing users to internal clouds reduces deployment times from days to minutes. Those shifts accelerate experimentation and business innovation, both critically important to competitive advantage.

MODERNIZE, OR FALL BEHIND

Companies who fail to modernize their data centers risk falling behind competitors. Best-in-class enterprises centralize their data centers, they map data center performance onto business performance, and they use the power of IT to reduce the speed of deployment, avoid unplanned outages, and keep better tabs on operations.

Modern companies require modern data centers, but transforming existing operations requires senior management attention. Those inside the data center can’t make it happen. Only management can begin transforming data centers from cost centers into cost-reducing profit centers, and that’s a result that everyone can cheer.

Modernizing Enterprise Data Centers for Fun and Profit, Jonathan Koomey’s online class for managers about modernizing data center operations will be offered from October 5 through November 13, 2015, in collaboration with Data Center Dynamics and Heatspring.

This article first appeared on the DataCenter Dynamics website on September 9, 2015.

Originally posted on
Jonathan Koomey
Written by

Jonathan Koomey

Jonathan Koomey is a Research Fellow at the Steyer-Taylor Center for Energy Policy and Finance at Stanford University. He has also held visiting professorships at Yale University (Fall 2009), Stanford University (2003-4 and Fall 2008), and the University of California, Berkeley’s Energy and Resources Group (Fall 2011), and was a Lecturer in Management at Stanford’s Graduate School of Business in Spring 2013. For more than eleven years he led Lawrence Berkeley National Laboratory’s (LBNL’s) End-Use Forecasting group, which analyzes markets for efficient products and technologies for improving the energy and environmental aspects of those products. The group developed recommendations for policymakers at the U.S. Environmental Protection Agency and the U.S. Department of Energy on ways to promote energy efficiency and prevent pollution. Koomey is also a Research Affiliate of the Energy and Resources Group at the University of California, Berkeley, and serves on the Editorial Board of the journal Contemporary Economic Policy. Dr. Koomey holds M.S. and Ph.D. degrees from the Energy and Resources Group at the University of California at Berkeley, and a B.A. in History of Science from Harvard University. He is the author or coauthor of nine books and more than two hundred articles and reports on energy efficiency and supply-side power technologies, energy economics, energy policy, environmental externalities, and global climate change. He has also published extensively on critical thinking skills.

More posts by Jonathan