Roger D. Lang, Director, Marketing and Business Development, Cornell Theory Center

Processor speeds are doubling, 64-bit computing is gaining ground, memory limits are soaring, and all the while prices are continuing to decline in the Wintel paradigm. Look back five years and the results appear to be revolutionary.

But this is not how the big breakthroughs are going to arrive.

The exponential developments are coming from combining the fast-growing power of Intel processors in clusters and grids and then making that power easy to use. This technique is not new. It has been variously labeled supercomputing, parallel computing, grid or cluster computing, but for convenience we will simply refer to it as high-performance computing (HPC). The challenge in the past was that HPC had been difficult to use. It was expensive to deploy, difficult to make all the processors work together, and challenging to write applications that could make good use of the clustered power.

Now, Cornell University, through its CTC Centers in Manhattan and Ithaca, is working with Microsoft and other vendors to achieve “off-the-shelf HPC”. Microsoft tools for the “Compute Cluster Edition” stack will enable solutions vendors to easily make their code HPC-ready. In fact, several pilot projects have already been deployed at major financial institutions and hedge funds, and the results are applications that scale nearly linearly. As it becomes easier to use and less expensive, HPC will move from the lab to much broader enterprise and retail markets where it will provide users with a dramatic competitive advantage.

In the past, supercomputing has been used for the most complex tasks, preferably when the federal government was footing the bill and price was no object – nuclear weapon design or weather modeling. In civilian research, HPC has been employed in genetic sequencing or petroleum engineering – lucrative businesses where no expense was to be spared on specialized technology.

Many of the challenges facing financial institutions are similarly complex, which explains why so many Ph.D.s from federal laboratories have moved to Wall Street with their advanced mathematical skills. However margins are thinner and it simply isn’t possible to allocate a Cray Supercomputer to every trader’s desktop. Now, out-of-the-box HPC will provide them, as well as smaller firms such as hedge funds, with the computational power they need to apply numerically demanding approaches to applications such as risk management, portfolio optimization, real-time trading analytics, trading signals and new techniques in data mining. In addition, parallelization of algorithms or simple brute force techniques will enable rapid advances in voice recognition that will dramatically change the way we interact with computers and the Internet, bringing on the next generation of applications for an even broader range of end-users. Instead of asking our computers for specific data or analysis, we’ll be asking them for indications of interest and using unstructured learning techniques to discover new clusters of behavior.

For solutions vendors there is huge incentive to use this new approach to HPC to move beyond the traditional $8 billion market for technical computing to a range of enterprise applications that’s at least 10 times as large.

Getting to Solutions

Companies will buy HPC clusters when they provide solutions to business problems. Vendors who want to succeed in HPC will need to apply solutions marketing techniques targeted at specific verticals, including:

• Identifying business issues in each vertical and aligning HPC technology and services to address required functionality. For example Basel II will demand the combination of market and credit risk, requiring increased performance and scalability of analytics.
• Develop a value proposition defining a competitive advantage. At Cornell we combine deep financial services expertise with world-class leadership in HPC, making us unique. What’s your sweet spot?
• Complete the solution with internal products, services, tools and external channels partners. Nobody can do it all!
• Package enterprise-ready solutions for specific markets, including positioning briefs, architecture design, technical white papers, services road maps, pricing, delivery and support.
• Develop your marketing material including Web content, client presentations, collateral and demos. CTC has helped many clients optimize their solutions on clusters.
• Drive marketing and awareness campaigns including trade shows, advertising, articles, press, client events and thought leadership through professional associations and organizations like CTC!
• Drive field readiness including sales training, sales tools, partner events and pre-sales support.
• Work directly with clients to define pilot projects and drive “marquis wins”.

At the Cornell Theory Center at 55 Broad St. in New York’s Financial District, we work directly with solutions vendors, systems integrators and end users such as hedge funds and investment banks to optimize their HPC solutions marketing programs, demos and prototypes.

To learn more about how the Cornell Theory Center can help your firm adopt HPC for competitive advantage, contact Roger Lang: Rlang@tc.cornell.edu or call 212-363-2915 x13.