High Performance Computing Cluster (HPCC) Update
Summer 2002

This past spring has been a busy one for the Committee on Technical Computing (CTC). The CTC is made up or representatives from the Colleges of Engineering, Science, Arts & Letters, Mendoza College of Business Business, along with representatives from the Office of Information Technologies. This committee provides direction and makes decisions pertaining to the HPCC and also the "High-End" OIT clusters.

This past spring an e-mail was sent from CTC Chairman Dr. Mark Stadtherr to all University Faculty communicating the expiration of the lease on the University's SGI 32 processor Origin 2000 machine (medusa.hpcc.nd.edu). This letter solicited input regarding various options concerning future HPCC facilities. These results were presented to the CTC.

This spring a user survey was sent requesting input from all HPCC users. Results from this were presented to the CTC which allowed a better understanding of the HPCC usage and especially the software used.

Thanks to the many Faculty and HPCC users which provided input for these surveys.

Along with input from these surveys, the CTC requested vendor presentations from IBM, Sun, Dell and SGI. This input was used in determining the best solutions for upgrading the HPCC.

In summary: The CTC decided upon the following:

When the lease of the 32 processor Origin 2000 medusa.hpcc.nd.edu) expires in early September 2002, the machine will be returned to SGI.

An IBM 1300 Linux cluster would be purchased and installed this summer. The computer portion of the cluster would consist of 32 IBM x330 compute nodes. Each node has 2 processors, thus adding 64 processors to the HPCC. A major part of the HPCC infrastructure is the addition of a Cisco 6509 Gigabit switch. This switch will provide interprocessor communication and fast access to shared disk. A interactive front-end machine will be available for interactive use and development similar to the Sun & SGI interactive front ends.

A major upgrade to the HPCC Infrastructure includes that addition of a IBM FAStT500 storage server. This storage server is a SAN which provides 1.7 Terabytes (usable) fast shared filesystem support to the cluster through 4 IBM x342 servers using GPFS. This additional disk space will also be made available to all current HPCC machines via the network using new Gigabit network connections added to those machines.

In addition, an 8 processor Sun V880 utilizing 900 MHz copper processors will also be added.

More specific details of the new hardware (and pictures) will be available in the next week on the HPCC web site. http://www.nd.edu/~hpcc

HPCC OUTAGES:
It was originally hoped that these systems would be available for "Early User Test Period" July 1st, due to delays in ordering / shipping we are now targeting July 15th. Watch the "Message-of-the-Day" for updates.

In order to update hardware/software, we will be planning outages of the current HPCC - tentatively on July 4th. Watch the HPCC "message-of-the-day" for further details. This will also allow us to update the batch queuing system to the latest version in order to better support of the Linux cluster. This outage requires that no jobs be queued or running. We are requesting that only short jobs be run until then, any jobs running at that time will be killed.

Please note that IP numbers of all the HPCC equipment will change during this outage. The campus domain name server (DNS) will be updated on the morning of July 4th.

If this would cause MAJOR problems, please notify us as soon as possible by sending e-mail to hpcc@nd.edu

page modified 12/12/02