Saturday, February 28, 2009

First steps

For those unfamilar with the ITAP datacenter in the MATH building, the space is actually divided up into thre rooms, G109 (aka the front room), G190 (aka the back room), and B60 down one floor. G109 is about 3000 sq ft, and G190 is about 5000 sq ft in size. The division is there, because the original late 1960s building (containing G109) was expanded in the early 1980s, adding G190, B60 (another smaller datacenter in the basement, around 2500 sq ft), and a row of basement offices. B60 is currently occupied by some ITAP development and testing systems, and colocation space for other departments on campus, including the Statistics department and CERIAS.

The physical preparation for the new cluster actually started last August, with the retiring of our aging IBM SP (the last parts were installed in 2001), and continued with retiring the IBM p690 Regattas and Sun F6800 cluster.

In addition, ITAP found some space in one of the old residence halls to rent space to run the recycled clusters in, helping to continue to provide service to users of that cluster, which includes researchers from around the university, who for one reason or another, do not have the ability or desire to buy into our community cluster projects, like Steele or Coates.

Thursday, February 26, 2009


Hi, my name is Patrick, and I'll be your tour guide for the next few months. You will have a front row seat watching how Purdue's research datacenter in the Math-Science building changes leading up to, and following the installation of the main section of the Coates high performance computing cluster, which is being planned for a May to June timeframe.

Following the success at Purdue with RCAC's Steele cluster installation last year, RCAC is working on deploying another cluster this year, called "Coates", after the founder of Purdue's Engineering Computer Network. Based on the response we've had from faculty interested in buying into the cluster, we expect to install between 800-1200 computers, which will give us between 3200 and 4800 more cores to run compute jobs on.

In this blog, I'll be posting pictures and information on the preparation that goes into installing the new cluster. The first few posts will be things that have already happened, catching up to today, and then I hope to post an update once a week or so.Link