HP announced Wednesday that Purdue University is using its container-based HP Performance-Optimized Data Center (POD) to expand and speed its ability to deliver research projects.
Faced with the need to expand its research compute cluster, Purdue found its operations constrained by budget, power and space limitations. The university was particularly impressed with the speed-to-market capabilities offered by the HP POD.
"We provide the resources for world-leading research, and delaying this work while building a new data center facility simply wasn’t an option,” said John Campbell, associate vice president, Academic Technologies, Purdue University. “With the HP POD, we’ll deploy an entire new data center in a matter of months at a fraction of the cost of a traditional data center, while being able to support all of our current, as well as anticipated, research initiatives.”
Economy Pinches Research Computing
With a turbulent economy in recent years and the rules for endowments changing for universities, the demands of research departments have driven them to act more like an enterprise and show the quick return on investment as well as protecting that investment in the future.
With technology changing rapidly, the future needs of the research clusters are difficult to forecast, and the container allows the incremental, modular approach while also delivering the end product in a matter of weeks instead of months or years. The dynamics of the computers installed for research clusters is changing dramatically, allowing the university to get much greater computing power in a much small footprint.
With that change in mind, the price tag of a traditional data center is problematic, especially when technological innovations could change the requirements of the data center before the ROI is realized. Purdue's research cluster can now expand by adding additional containers and technology as the needs dictate (and with the power that is available).
Options ranging from colocation to trying to find space in existing buildings and campus data centers were explored before Purdue ultimately decided on the container solution. Vendors were evaluated and the HP POD was selected.
Purdue reserved space inside the campus power plant property to pour cement pads for the POD containers. The pads also had networking conduit installed to network future PODs as they are installed. Having the data center close to the power source will help eliminate the possibility for power transfer and capacity issues.
Doing Without a Generator
With the close proximity to a power source, and a second utility feed available from Duke Energy, Purdue opted to not supply generator power for the POD. A small UPS and modular chiller were provisioned to support the POD. Similar to other supercomputer data centers, UPS is supplied to only critical storage and networking equipment.
To permit Purdue’s faculty to conduct leading-edge research, including modeling climate change and designing next-generation nanoscale electronics, Purdue’s Rosen Center for Advanced Computing also is building a new supercomputer. “Rossmann” is composed of a 1,000-node HP Cluster Platform 4000 based on HP ProLiant DL165z G7 servers with dual 12-core AMD Opteron 6100 series processors.
The HP POD will help Purdue support faculty from aeronautics, agronomy, climate science communications, medicinal chemistry, molecular pharmacology, biology, engineering, physics, statistics and other academia.
For more on the HP POD, check out this DCK video in which HP"s Steve Cumings outlines HP's approach to the container's design, as well as the economics of containers versus traditional brick-and-mortar data centers. This video runs about 8 minutes, 30 seconds.