On September 9 and 10 Altair held their high-performance computing virtual summit. Altair is a company with a large footprint. In their own words, “Altair is a global technology company that provides software and cloud solutions in the areas of data analytics, product development, and high-performance computing (HPC).” Their virtual summit did not disappoint. It was well-run, professionally produced and delivered strong, relevant information across a broad range of topics. I will attempt to summarize the latest in enterprise computing according to Altair.
Day 1 of the event was a plenary session devoted to topics of general interest. After attending a few virtual events, there are typically several “rough spots” that need to be negotiated. One is production vs. spontaneity. Pre-recorded presentations are predictable and reliable. You typically don’t have words dropping out or pets walking across the frame. On the other hand, they lack the genuine, spontaneous and engaging quality of a live stream. I believe Altair struck a good balance here. Bill Nitzberg, CTO of Altair PBS Works, kicked things off with a live introduction. He provided the transitions between the pre-recorded sessions as well. He is an engaging and high-energy speaker, so this helped a lot.
After the first day of presentations, a live Q&A session was held with all speakers. This was a very engaging and enjoyable session.
So, what was discussed on day 1? It was quite a lineup:
- Keynote Session: Exascale Computing Project, Michael Heroux – Exascale Computing Project
- PBS Professional 2020.1 Intro, Bill Nitzberg – CTO, PBS Works, Altair (Bill had the interesting task of introducing himself)
- Hierarchical Scheduling for High-throughput Computing Workloads, Jeremie Bourdoncle – Senior Director, Enterprise Computing Core Development, Altair
- Cloud Bursting: State of the Arts GUIs for Seamless Scaling, Chris Townend – Director, Enterprise Computing, Altair, Ian Littlewood – Product Manager, Enterprise Computing Core Development, Altair
- Budgeting, Allocation Management and Workload Simulation, Ian Littlewood – Product Manager, Enterprise Computing Core Development, Altair
On day 2 of the event, there were two parallel tracks. One focused on high-performance computing and the other on semiconductors. Given the typical background of the SemiWiki audience, I’ll focus a bit on the semiconductor track.
Stuart Taylor – Director, Enterprise Computing Core Development, Altair presented two topics:
- Saving Serious Money with License-first Scheduling
- Design Flow Mapping and Optimization: The Tool Every VLSI Team Needs
Andrea Casotto – Chief Scientist, Enterprise Computing Core Development, Altair also presented:
- Bringing Cost as Close as Possible to Exact Demand with Rapid Scaling
Some of you may remember Andrea. He was the long-time president of Runtime Design Automation, which was acquired by Altair in 2017. Andrea presented an excellent talk on a very relevant topic for cloud computing, one that I’ve had some experience with. He began his presentation by saying, “If you’re going to buy resources from any cloud vendor WATCH YOUR MONEY.”
Andrea provided some examples of runaway spending on cloud resources. This is quite easy to do actually. The good and bad news about the cloud is that you can get any compute resource you ask for. Not only do you get it very quickly, but you have to pay for it as well…
Andrea began by advancing the concept of “cloud elasticity.” The technology from Altair that Andrea discussed is called Rapid Scaling. Here is an important quote:
Rapid Scaling Brings the Cost as Close as Possible to Demand
What followed was a compelling discussion about methods to “right size” compute resources for the workload. There are a lot of good ideas in this presentation, some are patented. If you want to achieve the most efficient and cost-effective use of the resources from your cloud vendor, I highly recommend you listen to Andrea’s presentation.
You can see all the presentations from the Altair HPC Virtual Summit 2020 here. Check it out and see the latest in enterprise computing according to Altair.