The professional division of Thomson Reuters, six months into a massive data center virtualization project using Cisco Systems Inc. gear, has already seen a payback on costs.
The project using Cisco's Unified Computing System (UCS) was initially expected to cost $300 million over five years, but it's saving the company money by allowing it to tamp down data center growth and avoid capital equipment outlays, said Christopher Crowhurst, vice president of architecture at the Thomson Reuters Professional Division in New York. That division includes decision and support tools and applications for legal, accounting and health care professions.
Crowhurst said deploying UCS has already reduced data center complexity and allows computing services to be deployed faster. With UCS, “the connection between the network and the server is simpler,” he said. “There's no decision when you connect a physical server as to what the server needs to do.”
UCS, announced by Cisco in March, has been in trial deployments by several Cisco customers, though not all have been named. Thomson Reuters expects the full rollout, which is running ahead of schedule, to take about two years, Crowhurst said.
Crowhurst spoke during a virtual 30-minute video interview open to analysts and reporters who registered with Cisco to gain access over the Web. It was conducted as part of its Cisco Live! user conference being held this week in San Francisco. Reporters asked questions in text chats; those questions were relayed to Crowhurst by a Cisco marketing executive. The sound and video quality were fairly low and appeared to be transmitted via a handheld camera operated by the executive.
Using UCS has reduced the server count for Thomson Reuters from as many as 30 servers to one, Crowhurst said. And it has “radically” reduced the amount of cables used in the data center. Fewer cables generally means fewer problem, Crowhurst said, since in his experience, 90% of data center problems stem from cabling failures.
A side benefit of UCS is management ease, which, in turn, reduced the need for data center supervisors and required less focus on input-output demands. “That alone introduces a whole world of virtual candidates,” he said.
During the set-up process, Crowhurst said there were “horrendous failures. But we've learned a lot and the failures are [in] our playbook. …We have achieved tremendous stability for an X86 architecture.”