My recent blog posts have focused on performance enhancements to CAE applications and especially the scalability of ISV applications. A key point of this discussion is that it takes team effort to achieve these results. You need production models from the user community, the cooperation of application developers and computational scientists, access to large compute resources, and HPC system expertise. In this blog, I want to highlight the critical role HPC centers play in providing resources and expertise to improve the performance and effectiveness of CAE simulation for manufacturing companies.
Leading HPC Centers
Several of the largest HPC systems in the world are located at national HPC research centers. These centers are often associated with “grand challenge” science problems such as climate simulation, molecular modeling and nuclear physics. They provide both the necessary infrastructure and expertise to enable these simulations. Another key role these HPC centers play is aiding commercial companies in the use of HPC simulation to improve their competitiveness in the world market.
Cray systems are installed at many HPC centers around the world, including the National Center for Supercomputing Applications (NCSA) at the University of Illinois, the High Performance Computing Center Stuttgart (HLRS), Oak Ridge National Laboratory (ORNL) and Edinburgh Parallel Computing Centre (EPCC) just to name a few. All of these centers have programs in place to work with commercial companies and increasingly, we see these centers enabling leading-edge HPC simulation. HPC centers have become essential to expanding the competitiveness of commercial companies.
NCSA Private Sector Program
An excellent example of this partnership between an HPC center and the industrial community is the Private Sector Program (PSP) at NCSA. The PSP has 26 industrial partners and its core mission is “to help its partner community gain a competitive edge.” (www.ncsa.illinois.edu/industry) These partners include companies such as Boeing, Caterpillar, GE and Rolls Royce.
A core component of what NCSA is able to provide is a powerful computational resource. The “Blue Waters” supercomputer at NCSA is a Cray system and one of the most powerful supercomputers in the world with over 350,000 compute cores. The HPC team at NCSA is able to leverage the compute power of Blue Waters, ISV relationships and the industrial partnerships to enable new levels of performance in CAE simulation (again, it’s a team effort).
A recent example of this industrial cooperation is a leading-edge LS-DYNA simulation from an aerospace company which was run on Blue Waters by Seid Koric, senior technical lead for industrial projects at NCSA. The simulation scaled out to 15,360 compute cores and dramatically improved the simulation turnaround. This level of scalability for LS-DYNA is unprecedented on x86 compute cores, especially considering the complexity of the model geometry with multiple nonlinearities and difficult contact conditions. It is impractical, if not impossible, for industrial companies to divert their production HPC systems to test this level of scalability. However, the combination of the Blue Water system, the NCSA technical team and recent enhancements to the LS-DYNA code demonstrates the full simulation capability and allows the industrial partner to build this technology into their design process.
Similar projects are underway in the CFD area led by Ahmed Taha, senior computational resource coordinator at NCSA, using both Fluent and STAR-CCM+. Fluent has been scaled on the Blue Waters supercomputer to 10,000 cores in collaboration with ANSYS, Cray, NCSA, and a Private Sector Program partner company. Pushing extreme physics with one of ANSYS’s key customers, the Fluent software development team collaborated with experts in physics simulation applications to achieve performance at new levels. Again, working with their industrial partners, the NCSA team was able to demonstrate new levels of performance and scalability. This enhanced performance enables larger, higher-fidelity models which provide greater insight in the design process.
The cooperation of the ISVs in these projects is noteworthy. The fact that they are willing to participate and provide technical support and a highly scalable version of their application is in indication of the value of these projects and HPC in general. There is a constant demand for more complex simulations and this investment in high-end simulation provides performance efficiencies whether running 100 jobs each using 100 cores or a single job using 10,000 cores.
Ansys appreciated the achievement which NCSA-PSP accomplished with regards to benchmarking Fluent on Blue Waters and agreed to provide NCSA with a new “big” license key of (100,000) seats to continue the Fluent benchmark using different models.
Again, it is the combination of the industrial partner, the ISV and supercomputer resources that make it possible to achieve new levels of performance for CAE simulation.
Leading HPC Centers
NCSA is a leader in HPC and an excellent example of how an HPC center transfers capabilities to their industrial partners. Cray is fortunate to have a number of HPC centers located across the United States and around the world with industrial programs that are enabling virtual prototyping with increased accuracy and design insight. The table below lists a few of these sites. The quantity and quality of leading-edge CAE simulations which have been accomplished at these HPC centers is amazing. They have become key partners for Cray in demonstrating the critical value of HPC to the industrial community.
Examples of HPC Centers with Industrial Outreach Programs
|Center Name||Location||System Name||URL for more information|
|NCSA||Univ of IL||Blue Waters||www.ncsa.illinois.edu/industry|
|EPCC||Univ of Edinburgh||Archer||www.epcc.ed.ac.uk/blog/2013/11/20/archer-next-national-hpc-service-academic-research|
|HLRN||Berlin & Hannover||Konrad||www.hlrn.de/home/view/Service|
|UITS||Univ of Indiana||Big Red II||http://rt.uits.iu.edu/bigred2/|
Whether a company is producing airplanes, automobiles or microprocessors, they are in extremely competitive markets. HPC is an essential tool in the design process and enables greater fidelity simulations and greater throughput within limited design schedules, resulting in shortened product development cycles, improved product quality and lower product costs. As industrial companies look to incorporate data analytics, cloud computing and other emerging technologies into their product development process, HPC centers will play a vital role in demonstrating the value and introducing these capabilities to the industrial community.
HPC centers are more than a resource to the research science community; they are also a key enabler for HPC and simulation technology in the industrial sector.