DHAHRAN: Last Tuesday, the Information Technology Center (ITC) at King Fahd University of Petroleum & Minerals (www.kfupm.edu.sa) hosted a High-Performance Computing Awareness Day for its students and faculty. The event included Advanced Computing Seminars with outside speakers from Saudi Aramco, IBM, Microsoft and Cisco and a special presentation by ITC staff about the University's recently completed High-Performance Computing Project.
High-Performance Computing (HPC) uses supercomputers or computer clusters to provide the computation power needed for advanced research, particularly in science and engineering. During the final months of 2008, ITC installed a High-Performance Computing (HPC) platform at the KFUPM campus. The HPC cluster is a 128-node facility (1024 cores) delivering 12 TeraFLOPS and running a collection of essential systems and applications software. It can support applications, which run on either Linux or Microsoft Windows. The dual boot configuration enables administrators to power on any particular server or group of servers within the HPC cluster depending on the application, which needs to be run.
“The high-performance computing platform that will be inaugurated today, we hope, will contribute in a major way to further improve the research profile of the university and our faculty,” said KFUPM's Rector Dr. Khaled S. Al-Sultan. “Advanced research is heavily dependent on simulation and computational analysis techniques. Many of these tasks require large amounts of raw computing power. In the past, research computational power was delivered by supercomputers. With the proliferation of the new architectures, many computational resources are now provided by high-performance computing (HPC) clusters. KFUPM has always been at the forefront of the adoption and creation of the latest tools and techniques for research. In keeping with the advances in technology, it was decided to provide a scalable high-performance HPC cluster to our faculty to perform advanced research.”
The Rector explained that in the past, university faculty faced challenges in conducting research. In our connected world, research efforts often move much more rapidly now than in the past. This meant that the KFUPM faculty needed to reduce the amount of time they spent in testing various hypotheses and gathering data through modeling. KFUPM researchers also had to solve complex problems with large data sets. However, budgets are not unlimited at educational institutions. Achieving the best price for performance within limited funds allocation was essential for the HPC Project to be a success.
Today, HPC cluster resources are considered a core part of the global research infrastructure. Many laboratories and universities have such facilities including the NASA Ames Research Center, the Oak Ridge National Laboratory, the University of Tokyo, Sweden’s Umea University, and the National Center for Supercomputing applications at the University of Illinois.
For KFUPM, the availability of advanced computing resources is essential not only in helping faculty in advanced research but also in attracting top faculty, talented students and research grants. Before installation of the HPC platform, only low-end dual processor servers on the Windows/Intel and Unix-AIX/RISC platforms were available to the KFUPM research community. In the last few years, as the university invested heavily in providing advanced computing resources and infrastructure across the KFUPM campus, it became apparent that there was a need to furnish HPC resources that could be used for a variety of research.
According to KFUPM’s ITC Director Dr. Sadiq M. Sait, the HPC initiative received a major boost, when early in 2008, in response to faculty requirements and encouragement from the Rector, a plan was developed at ITC to enhance the current facilities. A university team was formed to handle the HPC project, including Dr. Sait, Tariq Maghrabi, Nabil Al-Halmoushi, Jaweed Yazdani, Farhan Khan, Khawar Khan and Khaja Asimuddin. As part of this effort, a detailed survey was initiated to study and analyze the requirements of KFUPM researchers.
“We did a survey to determine whether the cluster should run Linux or Windows. The results came back nearly equally divided. Some faculty wanted us to break the cluster into two and give half for Windows and half for Linux. That didn’t make sense if we wanted a powerful resource,” explained Dr. Sait. “ITC is very focused on providing the university community with its requirements, so we formed a small team to investigate how to do both operating systems. Now with our HPC we can do both and we have the only HPC resource in the region that has such a capability.”
The procurement process for the HPC project began in June 2008. In October 2008, the infrastructure for enhanced power and cooling requirements were deployed. Then for the past two months there have been intense efforts by the HPC team in the installation and testing of the HPC cluster. A number of applications such as Gaussian, DLPOLY and ANSYS were tested on the cluster setup. Test problems were sought by the team from research faculty in various departments including Chemistry, Physics and Mechanical Engineering. More applications are currently under testing and deployment on the cluster including MATLAB Distributed, FLUENT and GAMESS-UK.
“Definitely simulations and computations that would run on the order of weeks and months can now run in hours with the HPC cluster,” said Dr. Sait. “Things faculty could never think of doing before, now can be done. Larger test cases and more realistic test cases, now can be tried.”
Where is all this leading? The goal is to stimulate research at KFUPM. There has been frequent criticism that in Saudi Arabia and the region in general, there is not enough investment in R&D. The HPC cluster is an excellent resource, that when combined with human creativity, should bring new energy to Saudi R&D initiatives. A reception center for the HPC cluster has been set up at the ITC and a small HPC users group is already sharing ideas.
Superb to highlight as well, is that not only did the HPC project finish on time, but it went live at less than half the original budget.
“When we started last year, the HPC project budget was estimated at four million riyals for 128 processors. I am proud that the ITC HPC team launched the cluster for one point seven million riyals with double the original planned number of processors and even procured the most recent ones,” commented Dr. Sait. “We did our best to ensure that our faculty and students enjoy the benefits of parallelism, while keeping the option open to consider upgrades to the HPC platform in the future. Return on investment is very important to us. We keep looking at the road maps, revising our plans rapidly and working with vendors to get the best value. Admittedly, our HPC cluster is not as large as what is being set up at King Abdullah University of Science and Technology (KAUST), but it is what our faculty needs. We estimate that 90 percent of the research that would be considered by our faculty can be done with the new HPC. The plan is to coordinate with KAUST to book time on their supercomputer once it goes live to handle the remaining ten percent of projects that our HPC cluster can’t support.”
The installation of an HPC cluster at KFUPM will be only the second such resource available in Saudi Arabia. The other HPC cluster is at Saudi Aramco. The high-performance computing platform at KFUPM will provide a facility for high-level scientific research and also strengthen the position of the university as a premier academic and research institution in the Middle East. The facility will position the university to take up the challenges brought about by increased focus for localized high-level research in petrochemical and related industries.
“With all the research initiatives and incentives at KFUPM, with the opening of centers of excellences, investments in a nanotechnology laboratory, and with the continued quest to promote and excel in research, this activity is both timely and well aligned with the vision of the university,” said Dr. Al-Sultan. “Research in various areas ranging from Computational Fluid Dynamics, Simulation and Modeling, Seismic Tomography, Nano Sciences, Visualization, Weather Forecasting, Protein/Compound Synthesis, and on large data sets will be possible. In brief, you will be able to do things you could not do before. We hope to see a rise in research productivity in the near future.”
Comments to: [email protected]
The ‘Falcon’ will rise
Right now, for university researchers in Saudi Arabia, KFUPM's new HPC cluster is a dream come true. Eventually though, the Kingdom will have a more powerful resource at KAUST. At opening, KAUST's campus in Thuwal will house the Shaheen Supercomputer, a 16-rack IBM Blue Gene/P System, equipped with 4 Gbytes of memory per node and capable of 222 teraFLOPS - or 222 trillion floating-point-operations - per second.
High-performance storage and networking subsystems will support the capability of Shaheen. A 128 Linux node x86 Cluster will be provided as an auxiliary computational resource for pre- and post-processing, as well as initial x86 code tests prior to their enablement on Shaheen.
It is expected that eventually Shaheen will be available to other universities across the Kingdom on an established need basis. But Shaheen isn't scheduled to go live for another one to two years. This means that for the foreseeable future, Saudi academics will be able to look only to HPC@KFUPM to bring a big bang to their simulations. Even after Shaheen goes live, KFUPM's HPC cluster will still be essential to many researchers. The university plans to run a wide variety of software applications which will be of interest to a broader range of disciplines than the specialist applications which will be available on KAUST's Shaheen.
