INDUSTRY

By Tim Little HECMS -- The Oklahoma University 2006 Supercomputing Symposium held October 3-4, 2006 in Norman OK., started with a record number of attendees and vendors. With nearly 500 attendees and 15 HPC vendors, Dr. Stephen Wheat from Intel Corp. at the opening reception, stated this is the most important HPC conference in the central United States. In its fifth year OU has hosted this conference and it continues to grow each year. Dr. Henry Neeman, Director of OU’s Supercomputer Center for Education and Research (OSCER), believes the conference is a direct reflection of the need for broader education in computer science advancement. OU’s largest supercomputer named Topdawg, is a combination of Linux OS, Intel processors and LSF scheduler. Topdawg, currently 84th on the Top500 and 10th among universities supports a large user base which continues to demonstrate a need for more cycles. During the conference’s opening reception, Dr. Wheat stated Intel values the openness of the HPC market. He believes the ability to listen to the customer base and return this information to the corporate team has lead to great gains HPC. Intel is firm in its plan to make HPC a profitable market just as any other of its market segments. After the opening reception about 30 attendees had the opportunity for a personalized tour of the OU’s newest data center housing Topdawg. Henry Neeman was at his best and we all enjoyed seeing and hearing about the OU’s newest data center. OU’s largest supercomputer named Topdawg, is housed in the newest data and is combination of Linux OS, Intel processors and LSF scheduler. Topdawg, currently 84th on the Top500 and 10th among universities supports a large user base which continues to demonstrate a need for more cycles. The opening presentation at the conference was given by Dr. Henry Neeman. Dr. Neeman clearly stated having a fast computer is important but is not the focus of OU. During his presentation Neeman’s intentions were clear and driven. Educating people here, there and anywhere how to use computers to solve complex problems is certainly more important then having the fastest computer. In his own words, Neeman stated a supercomputer may last 3 years, and an education lasts forever. The keynote presentation was given by Dr. Daniel Atkins, Director, Office of Cyberinfrastructure, National Science Foundation - Dr. Atkins was quick to recognize Dr. Neeman for his exceptional and outstanding effort in the area of computer education. The NSF believes OU is one of the leaders in this area and is well aware of OU’s commitment to education in the Sciences. The NSF has been a long time supporter of OU activities and sees these types of investments paying off in the overall strengthening of the US’s technical advantage. Responding to the next generation of computing, the NFS is seeking ways to keep educational pace with this rapidly developing environment. The advancement in the speed of computing has a double-doubling effect. Computers are doubling in speed and the ability to practice more sciences that were not feasible just a few years ago is now reality. The NSF is addressing how to keep pace with this scientific advancement through global education and locating and supporting the best emerging technologies. The next speaker was Addison Snell, Director of HPC Research at IDC. Mr. Snell went over some business trends tracked by IDC. IDC projects continued growth within the HPC market and is seeing some dispersion in the Linux market. Comparing Linux to Unix, IDC feels there will be continued disjointed growth in Linux as many organizations develop there own version of this operation system. Today there is no good way for Independent Software Vendors to possibly keep current with all the variations of Linux. Then Mr. Snell went on to say that government spending as in the recent awards at Louisiana University and Texas Advanced Computing Center only help the vendor and institution and really to not benefit the overall market. Dr. Atkins from the NSF did not take well to this statement and asked for clarification. Mr. Snell then retracted his statement. Next up was Dr. Miron Livny Professor, University of Wisconsin. Dr. Livny who has received NSF funding for 20 some years for computer development and educational purposes talked in detail about CONDOR the open computing package developed at UW. In 1996, Livny coined the term High Throughput Computing (HTC). The goal of HTC was not to construct one large computer but to get the computing work completed in the most efficient manner possible. Through the development of CONDOR which is a computer sharing program the work is spread across a number of computer systems to enabling large workloads to be completed without owning or having to maintain a massive system or data center. The final Plenary Speaker was Dr. Stephen Wheat from Intel. Drug interaction processing is one of the leading fields Intel is tracking for HPC growth. The massive data base runs which are allowing pharmaceutical companies to model drug interaction with the human body requires a substantial a amount of computational speed and IO movement. Intel which has a strong roadmap and an even stronger commitment to HPC looked for and found ways to optimize these computing needs. With the advent of smaller and smaller wafers, Intel now has the ability to custom build operational specific dies along side its more traditional wafer. By combining mass production costs with specific die production, Intel can significantly reduce in the production cost of HPC specific processors. Intel’s internal research programs can see ways for the next twenty five years to push out more performance and contain the heat generation issues.