We were last in Cisco's new data center in Allen, Texas, in the fall of 2010 when the company was just putting the finishing touches on the 160,000 square foot building with 35,000 square feet of "raised floor" (they still use that lingo even though this facility doesn't use raised floors).
This data center, the crown jewel in the company's far reaching Global Data Center Strategy to consolidate and modernize core facilities, was brought online July 7, 2011 and we recently stopped back for an update (see our in-depth tour of the site under construction, or in pictures).
The Allen data center plays a critical role in the company's Cisco IT Elastic Infrastructure Services (CITEIS) private cloud, and is paired with a data center in Richardson, Texas, using Cisco's Metro Virtual Data Center architecture. MVDC enables one center to provide coverage for the other key applications, a fail-safe approach Cisco is using to safeguard critical applications.
Allen is 25% built out today, says James Cribari, Manager, Information Technology Global Infrastructure Services. "When we first opened, we had seven clusters of UCS (which is about 280 blades, 500 VMs), our installed raw storage was 1.5 petabytes, and our UPS critical load was 9%. Now it's 30% and we've grown to 45 clusters (1,192 blades, 4,000 VMs), and we have more than five petabytes of raw storage."
As built, the facility can house up to 250 UCS clusters, but that number is likely to climb as Cisco ramps up server density as new technologies emerge.
Many of the original data center design expectations are delivering as promised. The air side economizer, for example, was supposed to reduce the need for air conditioning by using outside air to cool the center 51% of the time, saving some $600,000 in electricity costs. So far, the outside temperature has allowed them to use ambient air 56% of the time, Cribari says. (They keep the data center at 78 degrees.)
The Power Usage Effectiveness (PUE) goal for the data center was 1.35, and that's better than what they have achieved to date, but to be expected. "We're probably sitting somewhere around 1.55 PUE today, but that's because we're not at the load that would be the optimum from an energy standpoint," Cribari says.
And the use of MVDC to pair Allen and Richardson to safeguard core applications is also going well, he says.
Not all applications, of course, need MVDC support, so the first step in rolling out that technology was to classify the criticality of the company's 1,500 applications, says Kirti Thakkar, a Cisco Information Technology Engineer.
They came up with five classifications, C1 to C5, with C1 requiring 99.999% availability (no acceptable downtime), C2 requiring 99.995% availability (with acceptable recovery time up to an hour), C3 requiring 99.99% availability (4 hours acceptable recovery time), C4 requiring 99.9% availability (24 hours of recovery time), and C5 with 99.9% availability and best effort recovery.
Sign up for Computerworld eNewsletters.