Cancel

Done

The University of Kentucky text as a link to the UK homepage.
The IT text as a link to the IT homepage.
Send A Message Button
IT Directory

New IBM Supercomputer Offers a Big Performance Jump

Jerry Grooms

Jerry Grooms, UK Systems Administrator with one blade from UK's new IBM System 1350

In just a few weeks research professors at the University of Kentucky will be able to process research computations much faster. UK has installed a new IBM Supercomputer offering a big performance jump. "The new lineup of processor systems and blades, high-speed connectivity and coprocessor acceleration technology for the System 1350 will enable researchers to do their computations about four to five times faster," says John Tibe, UK IT Director.

Making it even more complicated, in research, the movement is away from the symmetrical multiprocessor machines (SMPs) more towards the clusters. A computer cluster is a group of computers networked together and operating as a single computer-typically more cost-effective than a single computer of comparable speed. But some research codes still require SMPs machines. So UK must supply both.

Talk about a complicated juggling act!

Jerry Grooms and Matt Simpson, UK Systems Administrators, and Allan Hetzel and Butch Adkins, Data Center Operations, work in a small space--UK's Computer Center in McVey Hall. While maintaining the correct temperature, power, and necessary data backup for the existing High Performance Computing systems, they installed UK's new SMP and Linux Cluster systems from IBM.

To accommodate the installation of the IBM supercomputer, Allan Hetzel led the implementation of a Data Center plan that required over 20 steps:

  • relocation of six servers
  • relocation of four research clusters
  • relocation of three storage systems
  • relocation of an IBM blade center
  • installation of 36 60A power circuits
  • completion of five facility upgrades
  • 70+ hour work week to coordinate the installation of the ten racks of equipment

Since the first of January 2007, the new Cluster and SMP hardware has been put in place and connected, and the software installed. Next, the current research directories and data need to be ported over to the new clusters. This "juggling act" must be completed so the old SMP system can be removed.

In full production, the new IBM supercomputer:

  • requires over 130,000 W of power
  • requires over 400,000 BTU of cooling
  • requires 200 sqft of raised floor space
  • weighs approximately 14,000 lbs. (including 336 InfiniBand cables that weigh a combined 2,500 lbs.)

John Tibe

"UK is installing a new IBM Supercomputer offering a big performance jump," says John Tibe, UK IT Director

"With a fast and powerful supercomputer we can build, inside the computer, anything that we can imagine," says Gary Ferland, Physics professor and chairman of the High Performance Computing (HPC) selection committee.

"In my own work we apply the laws of physics to look at the events that occurred when galaxies and stars first started to form thirteen billion years ago. With the greater power these new IBM machines make available we can probe the birth of galaxies and stars more carefully than ever before." Like several scientific researchers, Ferland has his own computational code and programs which he runs for his research in Physics. Some chemistry and physics research is run on code generated by a an outside vendor.

Measuring Success

"One measurement of success is in Tera FLOPS--how UK's HPC stacks up against HPC's in other public institutions," says Doyle Friskney, UK IT Associate VP.

FLOPS (or flops) is an acronym meaning FLoating point Operations Per Second. This is used as a measure of a computer's performance, especially in fields of scientific calculations that make heavy use of floating point calculations; similar to instructions per second.

All of this will result in UK's research projects having faster computer computation available.