Dresden, Germany, June 28, 2006 – While most of those traveling to Germany this month are going for the 2006 FIFA World Cup soccer tournament, more than 1,000 academics and business professionals are heading to Germany for a different playing field: supercomputing.
Today is the opening day of the 2006 International Supercomputer Conference, the leading supercomputing event in Europe. One of the highlights of the event is the release of the semi-annual TOP500 project, an assembled list of the 500 most powerful, publicly known computer systems in the world. Started in 1993, the project provides a reliable basis for tracking and detecting trends in high-performance computing (HPC).
With the release of Microsoft Windows Compute Cluster Server 2003 at Microsoft Tech•Ed 2006 earlier this month, Microsoft aims to make high-performance computing (HPC) cluster technology more mainstream by bringing the cost advantages, ease of use and partner ecosystem of the Microsoft Windows Server platform to commercial industry and the public sector. Through Microsoft’s collaboration with the HPC community and strategic partners, Windows Compute Cluster Server 2003 is designed to deliver a more mainstream means for engineers, scientists and researchers to solve scaled-out business and scientific computational problems.
Rob Pennington, Chief Technology Officer, Leader of NCSA’s Innovative Systems Laboratory
One such collaboration is with the National Center for Supercomputing Applications (NCSA) at the University of Illinois at Urbana-Champaign, where Windows Compute Cluster Server 2003 served as the underlying operating system for a new HPC cluster that recently achieved 4.1 trillion computations per second (teraflops) on 896, 64-bit Intel Xeon processors. This result, arrived at by using Dell PowerEdge 1855 blade servers, Cisco Topspin InfiniBand switches and Force10 Gigabit Ethernet (GigE) switches, was sufficient to place the system at 130th on the Top500 list. The cluster, named Lincoln, will serve strategic campus and state initiatives, with its peak performance approaching 6 teraflops.
PressPass spoke with Kyril Faenov, director of high-performance computing at Microsoft, and Rob Pennington, chief technology officer and leader of NCSA’s Innovative Systems Laboratory, to discuss the Top500 result, the new working relationship between Microsoft and NCSA and emerging trends in HPC.
PressPass: Why is NCSA’s Top500 result with Windows so significant?
Pennington: Lincoln is the largest Windows-based HPC cluster available to researchers, demonstrating that Windows Compute Cluster Server 2003 can scale well beyond division and department-based deployments. It is also worth noting that Lincoln is the first NCSA cluster to support dual booting of both Windows and Linux.
Faenov: With Windows Compute Cluster Server 2003, our aim is to expand HPC beyond traditional supercomputing centers toward departments and divisions. However, it’s important for us to engage the traditional supercomputing centers like NCSA to understand the engineering challenges involved in putting together world-class HPC installations and demonstrate the headroom available with Windows HPC systems. NCSA’s installation is the largest and fastest benchmarked Windows-based HPC system to date, improving upon last year’s entry of two teraflops running on 660 processors and Gigabit Ethernet from Cornell Theory Center.
PressPass: Were there any other interesting technical aspects to the Lincoln system and Top500 result?
Faenov: NCSA’s Lincoln system represents one of the largest uses of Automated Deployment Servers (ADS) feature of Windows Server 2003. We used ADS to deploy the nodes via multicast and then managed the cluster through Windows Compute Cluster Server’s management tools and user interface. We also deployed Microsoft Operations Manager to monitor detailed cluster performance information in a scalable fashion. This installation also featured the largest deployment to date of the OpenIB Windows Infiniband driver, developed by the OpenFabrics community.
PressPass: It’s interesting that NCSA’s Lincoln system is able to dual boot Linux and Windows. Was this a goal of the project?
Pennington: We did set this as a goal. Today, almost all parallel applications running on HPC clusters operate on Linux. However, we know that there are plenty of technical applications, especially among our private sector partners, that run Windows. And our private sector partners are familiar with Windows, which certainly lowers the adoption curve when they transition to HPC clusters running on Windows CCS. I’d expect a Windows offering for HPC clusters would be well received by a large majority of academics used to Windows workstations and IT professionals from the private sector.
Kyril Faenov, Director of High-Performance Computing at Microsoft
Faenov: We are well aware that high-performance computing is heterogeneous and multi-faceted, with customers employing a breadth of resources, and different system architectures and configurations present at different ends of the performance spectrum. In addition to enabling partners to build products that run on Windows Compute Cluster Server 2003, we’ve been careful to invest in interoperability.
For example, we have a great partnership with Platform Computing – the leading vendor for job schedulers – and have done the work so our respective job schedulers can communicate with one another. Windows Compute Cluster Server 2003 also comes with the Microsoft Message Passing Interface, based on the reference implementation of the MPI2 standard, to make it easier to port existing parallel applications.
Also, from a development perspective, Visual Studio 2005 includes OpenMP support and a parallel debugger in order to support HPC application development.
PressPass: Tell us more about the new working relationship between Microsoft and NCSA.
Pennington: The Lincoln cluster heralds Microsoft’s entry into NCSA’s Private Sector Program, which is the gateway between NCSA’s research scientists and programmers and leading-edge firms interested in work being done at NCSA. This relationship will allow NCSA to remain on top of developments Microsoft is making in high-performance computing. Collectively, NCSA and Microsoft will address the interests of NCSA’s other private sector research and computing partners.
Faenov: Back in November, we announced a multiyear, multimillion-dollar investment in joint projects at 10 academic institutions around the world. The Microsoft Institutes for High-Performance Computing help guide ongoing software research and product innovation at Microsoft to address the most challenging technical computing problems.
Today, NCSA is becoming the eleventh Microsoft Institute for High-Performance Computing. Over the past 20 years, NCSA has made significant contributions to the birth and growth of the worldwide Internet-based infrastructure for science and engineering. And like Microsoft, NCSA is recognized as an international leader in its field.
PressPass: How does this working relationship benefit the HPC community?
Pennington: The work should significantly increase the opportunity for commercial and technical applications to tap into high-performance computing. We understand that there are technical applications, many of interest to our private sector partners, which run Windows. We expect that the need for HPC among these applications is going to be huge and it’s an important area to explore. These sorts of explorations are a key strength and mandate of NCSA’s Innovative Systems Laboratory.
Faenov: While developing Windows Compute Cluster Server 2003, we repeatedly heard that the HPC clusters available today are too costly and complex—both major barriers to mainstream customer adoption. So we’ve designed Windows Compute Cluster Server 2003 to be easier to deploy, use and manage, be it a departmental cluster used for Monte Carlo simulations or a large cluster like NCSA’s Lincoln used for more extreme computations.
And as Rob said, our work with NCSA and the other Institutes will help broaden the market for HPC applications. We’ve also partnered with strategic vendors who offer solutions for HPC, to include leading application vendors, OEMs and independent hardware vendors.
In terms of packaged applications, our goal is to have the top two or three applications in industries such as manufacturing, oil and gas, biosciences and finance ported to Windows. Today there are several applications available in each industry, and we expect that more will be available by the end of 2006.
