Humana Builds A Twenty-First Century Data Center For Healthcare

Healthcare costs face national scrutiny--and healthcare companies are responding with green data centers designed to deliver lower capital and operating expenses (CAPEX and OPEX).They are doing  this  by systematically reducing energy costs and by virtualizing away hardware and the costly software licensing fees that go with it. A case in point is Louisville, Kentucky-based Humana, Inc., a Fortune 200 corporation that employs 23,500 associates to market its health benefit consumer services in 50 states, Washington, D.C. and Puerto Rico. Humana decided in the mid-2000s to proceed with a green data center build that would not only deliver hard dollar savings over the long haul, but also be well-positioned for a business that has nearly tripled to 11.5 million participants over the past few years, making Humana the second largest Medicare prescription drug carrier in the United States and the most rapidly growing U.S. Medicare Advantage carrier.  

The data center challenge was evident to both executive and IT management, who recognized that successful business expansion depended upon technology that could support the needs of the business. Two Humana goals were a state-of the-art data center that would deliver scalability, flexibility and cost efficiency for information technology assets, as well as a focus on technology infrastructure to see what direct technology solutions could be brought to bear on reducing equipment power consumption, footprints, costs and overall IT labor costs.

Humana's green data center was replacing a 60-year-old facility. "We wanted flexibility,  cost efficiency, expandability and the ability to respond to any emerging technology need resulting from business evolution and growth," said Alan Ainsworth, Humana's Director of IT Operations.

Humana and its vendor/consultant data center team immediately took aim at PUE (power usage effectiveness) in the design phase of the new facility. PUE is a metric used to determine the energy efficiency of a data center. It does this by dividing the amount of power entering the data center by the power used to run the data center's IT infrastructure. As overall data center efficiency increases, the PUE quotient, or ratio approaches 1. "A perfect PUE ratio is a 1, but that is virtually impossible to get to," said Ainsworth.

The team's data center modeling incorporated requirements for floor and ceiling designs, power generation and distribution, cooling, rack specifications and network cable distribution.  In the course of each modeling exercise, different room configurations, equipment placements, air conditioning locations and sizes, under-floor cold air efficiency and hot air evacuation designs were evaluated. "Several threads of our project centered around best practices, such as the HVAC (heat, ventilation, air conditioning) and power cooling areas," said Ainsworth.Every modeling exercise showed how the under-floor static pressure distribution, power consumption, rack temperatures and air flows would react under different configurations. In turn, these models were plugged into data center usage forecasts three years out and also into models for the data center that would be at both full and maximum build-out.  

Evaluating IT Infrastructure
Humana next evaluated IT infrastructure to see where equipment footprints and energy consumption could be reduced. Humana had two IBM mainframe System z10s running its claims and billing systems. These were not causative factors in server sprawl, and they also handled the heart of  Humana's mission-critical processing. Where Humana was experiencing server sprawl was with its Intel servers. "Our storage and our server population rapidly expanded as a consequence of a significant period of growth in 2005-2006 because of the Medicare Part D drug initiative," said Ainsworth.

To improve storage, Humana converted from Sun 9840C to Sun 9840D tape drives, effectively doubling the capacity of tape cartridges. On the server side, Humana worked with Intel Solution Services by virtualizing physical servers with VMware. Multi-core server platforms were configured in various combinations with blade centers, setting the stage for flexible, virtualized computing that significantly affected the heat load characteristics of the data center and also reduced IT infrastructure loads. The result was scalability that now allows Humana to purchase new equipment, whether servers, storage or air conditioning, in a just-in-time (JIT) mode, never before it is needed.

 "This is still a project "in process," and we have not yet realized all of the dollar savings we expect to see," said Kelly Vogt, Manager of IT Operations for the Louisville Data Center. "This is because we are still virtualizing Windows boxes every day. Virtualization is improving the results daily, and we anticipate a significant savings in data center power consumption, coupled with an extension of the useful life of the data center itself."
 
With 1,900 servers in its new data center, the potential to virtualize some of these servers on the System z mainframes is also on Humana's radar. "We approached initial phases of this project by first using virtualization and improving our server densities on the open systems side of our operation," said Ainsworth."Now we are starting to look at zLinux virtualization, and the possibility of taking advantage of some attractive cost savings IBM makes available by using zIIP (System z Integrated Information Processor) engines."

The construction of Humana's data new data center, and the initial phases of revisions to its IT infrastructure to render computing more cost and energy efficient, were resoundingly successful, but like any project of this magnitude, challenges surfaced along the way that Humana IT  had to contend with and resolve."At the height of the project, a couple of hundred people were involved," said Vogt. "Before starting anything, we put in a significant amount of planning and created an overall interdisciplinary team comprised of various sub-teams with their own project managers. We had an overall Program Office for the data center migration project that was led by two people who had all of the operational and applications resources available to them. We engaged consultants where needed, and included virtually every vendor that had hardware, software or other equipment in the data center."

Ainsworth projects that the new data center, which is at around two-thirds capacity, will improve its power usage effectiveness ratio as times goes by. "Right now, we are at the 1.6-1.7 ratio level for PUE," said Ainsworth, "But because we have excess capacity today, that ratio is higher. As we build out data center IT resource utilization, I expect to see that metric go down toward the 1.0 PUE that we strive for."