Just two weeks ago, the high-performance computing (HPC) industry converged in Denver for the biggest event of the year – SC19. It was a successful show with broad representation from across the industry. Attendees soaked up expert insight, participated in birds-of-a-feather sessions, and took advantage of excellent networking opportunities.
Our booth saw high traffic and our theater presentation crowds grew to overflow the seating capacity. Our speakers represented more than a dozen customers including Lockheed Martin, Northup Grumman, 3M, and more. Altair experts visited many of our partner booths and presented to crowds gathered for Intel, AMD, SUSE, and several of our cloud partners.
Exascale and the Aurora Supercomputer
There was no better place than SC19 to announce our partnership with Argonne National Laboratory, in support of the pursuit of exascale with the Cray-designed Aurora system that is slated for delivery in 2021. Argonne joined us for an exciting live Q&A on exascale job scheduling, and in the next years we look forward to helping the lab team reach their goal.
Aurora will be capable of performing a quintillion calculations per second and the system will be in high demand from researchers around the world, so effective workload management will be crucial.
Talking exascale with Bill Allcock from Argonne National Laboratory and Altair’s Bill Nitzberg
Altair HPC Solutions
We love to share latest advances in Altair technology, and interest during the show was high. Hot topics included Altair Accelerator™ Plus and Altair PBS Professional™ for hybrid workload environments. A combination of these can handle both massively parallel jobs and the high-throughput, short-running, single-core jobs required by many commercial and research organizations.
Our experts shared results from testing solutions including SAGA (storage-aware grid acceleration), powered by Altair Accelerator on Dell Isilon architecture, which demonstrated significant speedups over Slurm. Altair chief scientist Andrea Casotto discussed Altair’s RapidScaling in the AWS cloud, and how to bring cloud spend closer to exact compute demand.
One of the most popular presentations at the Altair theater was, “Cloud Bursting for Modeling and Simulation Workloads,” by 3M. Our booth visitors appreciated hearing directly from a customer about how easy it is to leverage elastic cloud resources with Altair Control™.
3M cloud HPC engineer Gabe Turner presents on cloud bursting at the Altair theater
Our participation at SC is always a strong testament to our ecosystem of HPC hardware, software, and cloud partners. Through our collaboration with partners such as HPE, Arm, AMD, Oracle, NVIDIA, Microsoft, AWS, Intel, Lenovo, Dell-EMC, Atos, Google, and several others, we continue to amplify Altair’s position as a leader in simulation and solver technology.
This year we had an excellent showing, and Altair had more than two dozen partner engagements. We were invited to present at a variety of user group meetings, panels, birds-of-a-feather discussions, on-site theater presentations, and technical demos.
At the HP-CAST and Arm user group meetings, we shared Altair Radioss™ solver performance on the UK Catalyst Bristol cluster, which includes Arm-based HPE Apollo 70 systems. The presentation included details about the high efficiency achieved with Altair Radioss in testing different MPI libraries and setups. This was shown by running several industrial examples on a multi-node HPE cluster equipped with Marvell ThunderX2 Arm processors, which is a cost and energy-efficient option.
At the Intel theater, we presented with Microsoft on Radioss performance running in the Azure cloud on Intel instances. We were also invited to contribute to an Intel panel on simplifying HPC in the cloud. SC visitors also learned about our hybrid cloud bursting on the Google Cloud Platform, Oracle Cloud Infrastructure, and Microsoft Azure platform with Altair.
Additional partner presentations included AMD on the future of HPC with Altair Radioss and AMD EPYC™ processors, and Lenovo on powering complex engineering applications with Altair HyperWorks™ solvers, where we shared benchmark performance for Altair AcuSolve™, Altair Feko™ and Altair OptiStruct™. At the AWS theatre we shared findings, recommendations, and best practices for running AcuSolve and OptiStruct on AWS.
Altair took center stage twice at the NVIDIA theater, where we shared solutions for boosting GPU performance in a highly complex GPU environment, as well as GPU-specific uses cases including our work with Volkswagen Group Research on AWS.
For the EDA space, we shared the stage with Dell EMC at the Dell theater to describe how Altair Accelerator™ helps companies do more through storage-aware grid acceleration.
We participated in several related events in parallel with the conference, including a Tuesday night happy hour co-hosted by Oracle.
As the conference was wrapping up on Thursday, the Cylc user group learned about how the biggest weather sites in the world orchestrate cycling workflows with advanced HPC tools like open-source Cylc and PBS Professional. The half-day session included lunch and featured talks by Cylc founder Dr. Hilary Oliver from NIWA, as well as experts from Australia’s Bureau of Meteorology, the U.S. Naval Research Laboratory, and Altair.
As technology continues to advance and we move further down the road to exascale, we look forward to seeing what’s next and connecting with our customers, partners and prospects next year at SC20 in Atlanta.
Learn more about Altair’s HPC solutions, here.