At Interop, Best Practices for Data Center Standardization

By David Rubinstein

May 1, 2008 —  (Page 1 of 3)
LAS VEGAS—Standardization within the data center has to occur in both the physical infrastructure as well as the logical sense. Two speakers with very different sets of problems laid out what they called best practices for achieving uniformity.

Speaking about the logical side was Dave Leonard, CTO of an IT infrastructure outsourcing company called Infocrossing, which was acquired by Wipro in 2007. The company faced the problem of having five data centers, added via acquisition, that each had its own tools, sets of processes and team structures. Leonard said the first steps the company took were to break the ties to geography and technology, so any person could manage any device from any location.

Things working against standardization were the employees, who had their own fiefdoms and who also feared job loss through consolidation; access to the network and devices, along with security and rights management issues; the fact that knowledge was local; and the installation of different hardware infrastructures in the different locations.

But by getting the employees to buy in to the plan and empowering them to work at a higher level, their productivity and satisfaction went up, Leonard said. By standardizing the tools in use throughout the organization—after determining they were affordable, could do the job and the employees would actually use them—purchase and maintenance costs dropped significantly.

Meanwhile, Carl Cottuli, vice president of customer projects at hardware provider APC, focused on power, cooling and management standardization. The obvious benefits, he said, are reduced downtime, better business agility and decreased cost of ownership. However, there also are gains to be had by making the data center predictable, increase the amount of human learning, and relieve the confusion and fear of change.

Old data centers were custom-designed, made up of components from many suppliers and involved complex integrations to get the components to work together. This process was slow and error-prone, and led to many defects, Cottuli said. By contrast, today’s data centers are moving toward a modular row model, in which power, cooling and the management infrastructure are included in a self-contained unit that can be deployed anywhere in the data center on demand. “Unique data center designs created unique problems,” Cottuli said, that people could not learn from to anticipate future failures. “Standardization,” he added, “eliminates one-time engineering.”

Related Search Term(s): Data centers, power, storage hardware, systems management, testing & troubleshooting

Share this link: http://www.sysmannews.com/link/32120