DAILY NEWS AND INFORMATION FOR THE GLOBAL GRID COMMUNITY / JUNE 23, 2003: VOL. 2 NO. 25    ( Table of Contents )

Special Features:

EXCLUSIVE INTERVIEW: DELL LAYS OUT ITS HPCC STRATEGY
By Neil Alger, GRIDtoday Correspondent

GRIDtoday: What can Our readers expect to see from Dell at ClusterWorld?

Dr. Rooholamini: Dell is a sponsor of Clusterworld this year. We have been involved with the tradeshow organizers from its inception, providing feedback on formats, technical content and possible topics. We have also encouraged Dell customers to present their best practices in cluster computing in the technical track at Clusterworld. We want end users to hear testimonials from Dell's existing customers about how clustering has helped solve their computing problems.

Two universities who will present their research are the University of Texas at Austin and Cornell University. Dr. Kamy Sepehrnoori from the UT-Austin is working with Dell on the use of standardized clusters for petroleum research. He will present a paper on Parallel Reservoir Simulation on Dell Clusters - several of my team members collaborated with Dr. Sepehrnoori in this research and co-authored the paper. Cornell Theory Center has several Windows-based clusters running on Dell systems used in a variety of research including life sciences, finance, physics, and chemistry. Researcher Jaroslaw Pillardy is presenting a paper on the topic of Parallel Computational Biology Tools and Applications for Windows. Cornell is widely regarded on its development of tools for organizations that wish to use Windows clusters.

GRIDtoday: How is Dell differentiating itself in the rapidly expanding clustering market?

Dr. Rooholamini: Dell not only focuses on standards-based solutions that are scalable and modular, but also drives standardization as a value proposition to its customers. This provides our customers with the best possible price/performance and the flexibility to add power and capacity to their clusters as they need it. We estimate that Dell clusters cost approximately 10% of a proprietary platform and perform better for the majority of applications. We feel that Dell's focus on customer needs, best-of-class partnerships and simplifying the purchase, deployment and management of clusters are both a differentiator and a driver of expansion in the clustering market. Dell's ability to provide an end-to-end solution makes it a logical choice when purchasing clusters as the customer receives everything they need from one vendor. This provides for a single point of contact for sales, consulting, service, and support.

It is exciting to see clusters grow so rapidly and be adopted by national labs, universities and corporations. The collaboration that Dell has taken part in with universities has allowed us access to advanced technologies, which in turn, we have provided to our larger corporate customer base as product offerings. In the meantime, through this collaboration, Dell contributes to the research community and ensures future improvement in areas of mutual interest.

GRIDtoday: How would you define Dell's competition, and how is Dell positioning itself to meet the challenge?

Dr. Rooholamini: Dell provides standards-based solutions for high-performance cluster computing that offer the latest processing, storage, interconnect and management capabilities for the maximum value and flexibility available. Our low total cost of ownership is even more appealing in the cluster space due to the larger number of systems deployed per user. Dell's direct-model has made the preceding possible through its ability to bring customers, vendors, and technology collaborators together. We see these solutions meeting or outperforming proprietary systems, but costing 90% less than them. Through our HPCC program, we test and partner with best-of-class hardware and software companies to help simplify customers HPCC deployments and manage their growth. Customers have embraced this approach - we believe this is reflected in our No.1 market share position in the Linux Technical Clustering market according to IDC.

GRIDtoday: Over the last year, HPCC has really taken off, not only in academic and research environments, but in the enterprise arena as well. What do you think was the most important breakthrough in the cluster technology field over the last twelve months that has helped to facilitate this expansion?

Dr. Rooholamini: Standardization of technology has driven improvements in processor density, storage capacities, and interconnect performance. Over the last 12 months we have seen a dramatic reduction in price/performance as well. These improvements have all contributed to the large-scale adoption of HPCC in all customer areas. This has allowed for adoption of HPCC by customers that hadn't used them before, couldn't afford them, or had previously used proprietary systems. Today, more developers have access to clusters to create new parallel applications that can take advantage of the scale out capabilities of HPCC. In turn, these new applications make HPCC more attractive to other developers, researchers, and companies.

GRIDtoday: As HPCC migrates into the enterprise, what hurdles stand in the way of its uptake? What kinds of strategies is Dell employing to facilitate enterprise implementations of cluster computing?

Dr. Rooholamini: Availability of "parallelized" applications, ease of management, and ease of access are all areas of focus for HPCC in the enterprise. We work with ISVs to port, optimize and tune their applications for HPC, continue to evolve the Dell OpenManage for managing Dell hardware through in-band and out-of-band routes, and work with service providers to address the access and availability of computing resources at affordable prices.

GRIDtoday: As cluster computing becomes more ubiquitous in the academic research arena, how is the relationship between the researcher and the computer being altered?

Dr. Rooholamini: This is a very interesting question. Before the advent of HPCC most research computing was done on supercomputers where inconvenient and expensive time-sharing was required amongst a cadre of researchers. This resulted in long delays for jobs to be run and limited access to supercomputers. With the low cost and high-performance of relatively small clusters, we have effectively democratized supercomputing so that researchers and professors can now have powerful supercomputers in their offices if they'd like, or can have access to a pooled set of resources with multiple colleagues. I have always said I'd like to see a cluster in every classroom - think of the possibilities and amounts of research that could be conducted. We could be very close to this possibility.

GRIDtoday: At this point in time, does the reality of HPCC in the research environment live up to the hype?

Dr. Rooholamini: If by hype you mean that more and more researchers are building clusters to conduct their most advanced research and data analysis, then yes, it is living up to the hype and more. Dell customers are using their clusters to study the human genome, design new drugs, study cancer cells in hopes for a cure, track disease epidemics like SARS and small pox, fight bio-terrorism, explore space, determine the origins of the universe and more. The work being conducted on standardized Dell computers is truly inspiring.

GRIDtoday: Where does Dell see HPCC five years down the line in the research community? in the enterprise? in the home? What do you see as the single greatest hurdle facing cluster technology for reaching this vision?

Dr. Rooholamini: Regardless of environment, we see standardization of the hardware and software building blocks mature in the coming years. This will lead to the adoption of cluster computing as a paradigm to solve many of the "grand challenge" problems facing us today, as well as addressing some ordinary ones. We see the quality of the information that we base our decisions upon improved, whether it is in analyzing financial data, building a safer car, predicting the weather, sequencing genetic data, etc. We see "large" computing resources become accessible via grids and more affordable for enterprises to utilize. We hope to see the rate of scientific discovery accelerate and cure for diseases such as cancer and AIDS are found in the near future.

In realizing this vision, we do not see any major hurdles. In our mind, the scale-out model of solving problems, whether in technical computing, database serving, file and print serving, messaging, etc. has won out, and there will be more usage models based on this paradigm in all application domains in the years to come.

GRIDtoday: What can our readers expect from Dell in the coming months?

Dr. Rooholamini: Customers will continue to see Dell provide standards-based solutions that are powerful, robust and cost-effective for all customers. We will continue to partner to advance the reach of our hardware platforms and expect to announce many new customers in both corporate and academic environments. We feel that the true expansion of the market for clustering is yet to come as customers migrate to standards from their proprietary UNIX based systems.

GRIDtoday: Is there anything that you would like to add for our readers?

Dr. Rooholamini: I'd like to thank you readership for giving me this opportunity to speak with them. We believe in a direct relationship with our customers and see informed publications such as GRIDtoday to help us accomplish this important goal.

Dr. Rooholamini serves as director of Enterprise Solutions at Dell where he focuses on operating systems, clustering, and application development for servers. His product development responsibilities include Dell PowerEdge SMP servers and all cluster product offerings from Dell.

Before coming to Dell in 1994, he worked for Memorex Telex and AT&T where he was responsible for advanced client-server systems and networking products. Dr. Rooholamini was formerly a professor of Computer Systems at the University of Wisconsin and has over 16 years of experience in computer and communications industries. He has published over two dozen publications in conferences and journals.