GRIDtoday Logo Altair

DAILY NEWS AND INFORMATION FOR THE GLOBAL GRID COMMUNITY / JANUARY 20, 2003: VOL. 2 NO. 3

( Previous Article )   ( Table of Contents )   ( Next Article )

Special Features:

GRID-BASED SUPERCOMPUTING: INTERVIEW WITH IBM'S DAVID JURSIK
By Alan Beck, Editor-in-Chief

IBM recently introduced a Grid-based Supercomputing On Demand service where customers have the option to either buy POWER or Intel® processor-based supercomputer clusters or access these systems on demand, paying for processing power based on the required capacity and duration of use. To learn more, GRIDtoday interviewed David Jursik, IBM's VP of Linux Clusters.

GRIDtoday: What specific factors led to IBM's introduction of Supercomputing on demand?

JURSIK: Frequent customer requests for short term capacity to satisfy their needs to:

  • Complete functional application testing and migration

  • Perform application and system stress tests

  • Validate research findings and a variety of other fixed duration projects that all exhibited the requirement for large capacity supercomputing cycles, in a controlled, robust, data center environment, utilizing the latest technology with strong systems management disciplines, led IBM to realize the need for this type of service.

GRIDtoday: How does this initiative differ from prior Grid computing services offered by IBM?

JURSIK: IBM sees the usage and adoption of Grid computing technologies spreading rapidly and we expect that the variations of usages will be many over time. For example, we are working with one client to link their internal Grid to additional capacity from IBM which would work in concert with the client's internal Grid to handle requirements for peak capacity or for business critical tasks that require accelerated turn-around times.

GRIDtoday: Whom did you foresee as the typical Supercomputing on demand customers?

JURSIK: IBM believes that Petroleum clients performing seismic analysis, Life Sciences companies performing genomic studies and movie studios creating animation sequences are all examples of businesses with short term capacity requirements that have a well defined project length and a finite need for a specific amount of capacity.

GRIDtoday: How does IBM pricing compare with similar services offered by Gateway and Sun?

JURSIK: Pricing will differ from customer to customer based on various variables. We can't comment on Gateway or Sun's offerings.

GRIDtoday: How are you marketing Supercomputing on demand to potential end- users? How are such potential customers identified?

JURSIK: IBM is the world's leading provider of high performance computing solutions as demonstrated by our position on the TOP 500 supercomputing list. Our experiences in the technical computing market over the past 10-15 years has allowed us to better understand its requirements and in doing so, we have worked with many clients in the high potential application areas mentioned previously that have asked for these solutions. In short, our focus on solutions, across our services, software and hardware brands has led to a unique working relationship and understanding of the time dependent needs of many researchers and other supercomputing users and we have structured this offering to specifically meet those needs.

GRIDtoday: In Grid computation, there is often a significant difference between pure processing costs and expenses incurred in setting up a problem or function for processing. How will IBM tackle this dichotomy?

JURSIK: IBM is well positioned to provided consultative services to our customers in the definition of problem statements, the evaluation of alternatives, uses of appropriate tools and preparation for the execution of complex tasks. Our investments in the Grid Computing initiatives over the past several years has allowed IBM to apply the intellectual capital in helping our customers with the less resource intensive phases of their project. Complementing those skills, we will have large computational capacity available at the time required, for the in amount dictated by the problem definition and for the duration required to solve the problem. By combining both these consultative skills and the short term computational capability, we believe we are uniquely prepared to address these needs.

GRIDtoday: What were the most significant technical challenges involved in rolling out Supercomputing on Demand? How long was the development process?

JURSIK: It turns out that for the class of customers and applications we are targeting to serve in the initial stages of our roll-out the technology we have developed for our clustering business over many years is well suited to accommodate the needs of the customers we have engaged.

This includes management of large numbers of nodes, scheduling and workload balancing technology, monitoring capability, security, system scalability and so on. Thus, our ability to move quickly in this arena has been enhanced by virtue of the availability of technology many of our customers make use of today.

There are, of course, other technical challenges that will arise as we broaden the community of customers we serve and experience more diversity in the nature of the applications targeted. We expect that some of these issues will revolve around resource balancing that we expect to target with our Grid technologies which we further expect will enable us to tie together distributed facilities to amplify the amount and type of technology eligible to be deployed against customer demands.

GRIDtoday: How will the market niche for Grid computing evolve over the next year? The next decade? What will IBM's role be in this process?

JURSIK: IBM is working to drive the movement of Grid from its roots in academia and research into mainstream commercial computing. At this time Grid technologies are being adopted by technical "achievers" in early adopter segments such as financial services, automotive, petroleum, and life sciences. Our initial proof-of-value projects in these segments are demonstrating significant business value. These results will most certainly serve to accelerate the penetration in these segments.

Utilizing the introduction of Globus Toolkit 3.0 and the first versions of the Open Grid Services Architecture (OGSA) standard in 2003, we are also now working with key industry ISVs to "Grid enable" 3rd party applications. This will deliver the benefits of Grid, such as better access to data and increased productivity, to a much wider set of users.

http://www.ibm.com.

( Top of Page )

( Previous Article )   ( Table of Contents )   ( Next Article )