Whats missing from the 'missing middle'?

Share this on social media:


David Hudak thinks there is a lesson from the movie Soylent Green for anyone trying to encourage the uptake of HPC by manufacturing industry.

The primary focus of HPC has been modelling and simulation from the beginning – accurately simulating physical problems was the initial driver for high-performance computing. Since their inception, academic HPC centres such as, in the United States, the National Center for Supercomputing Applications, Texas Advanced Computing Center, San Diego Supercomputer Center and Ohio Supercomputer Center all had policies along the lines of: ‘if we can help industry, we will’. But for a long time, direct industry/HPC centre engagements were sporadic at best.

This has improved over the past ten years. Industry and academic HPC centres have built sustained collaborations. The companies are from a variety of sectors with some showing greater interest, such as energy, automobile and aerospace manufacturing, and biotechnology. Multiple public/private partnerships have been launched to address barriers to HPC adoption, including expertise, software, and access to hardware. I head the Industry Relations programme for the Extreme Science and Engineering Discovery Environment (XSEDE), which is an initiative of the US National Science Foundation. XSEDE is a single virtual system that scientists and engineers can use to interactively share computing resources, data and expertise. In my industry relations role, I enjoy talking about XSEDE’s part in addressing these barriers. For example, we are trying to address the ‘expertise barrier’ by leveraging our training and education programmes for workforce development.

So, are all the pieces in place? Are we on the precipice of a new era in modelling and simulation in support of industry, particularly manufacturing? The answers are ’no’ and ‘maybe’.

We are still searching for the ‘missing middle’. Again, around ten years ago, it was noted that only the largest Tier-1 manufacturers like Procter & Gamble used HPC-based modelling and simulation. It was argued then by many “serious people” (myself included) that small-to-medium manufacturers were not using HPC due to up-front costs of hardware, software, and expert analysis. This category of companies was referred to as the missing middle. The argument went that if the hardware and software became affordable, the missing middle would begin to adopt HPC. Well, a lot has changed in the past ten years. Academic HPC centres and commercial cloud providers alike now provide hardware on demand and at reasonable rates. Independent software vendors are evolving their licensing plans to accommodate short-term usage (albeit too slowly). Promising open-source packages like OpenFoam and WARP3D reduce software costs even further. Hardware and software are much more available. And yet, the middle is still missing.


I think the root of the missing middle is the fundamentally different view of HPC taken by manufacturers and HPC centres. Being from the academic community, I think of HPC as (1) install a big machine and (2) write some software for that machine. Manufacturers would rather not do either of those. Industry is not concerned about directly owning and operating HPC hardware if cheaper outsourcing alternatives exist. As Tom Wilkie noted in a recent SCW op-ed, ‘Airbus’s business depends on HPC… But HPC is not its business.’ Similarly, manufacturers are not interested in writing software. In my experience, manufacturers are interested in pre-packaged ‘Turbo Tax’ apps rather than full FEA or CFD solvers. (Disclosure: I participate in an initiative by the Ohio Supercomputer Center called AweSim that is developing a platform for creating, hosting, and selling such apps.)

But even if simulation apps exist, manufacturers must customise and integrate them into business processes before realising any value. To meet this need, Tier-1 companies internally develop ‘simulation technicians’ who understand the manufacturing processes and are trained to apply simulation apps. These practitioners are part of the manufacturing team and improve solutions, in part, by using the apps. These techs underscore the need for industry best practices on modelling and simulation adoption – when, where, and how to use HPC. HPC in and of itself is not a competitive advantage for industry. Improving a company’s capability to design, test, and manufacture products through HPC provides the competitive advantage. A key element for the missing middle is recognising ‘simulation techs’ play a unique role with their own expectations, skill sets, and training. These practitioners will improve the bottom-line through modelling and simulation. To paraphrase Charlton Heston from the end of “Soylent Green” (spoiler alert!), ‘The missing middle is made out of people!!!

David Hudak serves as the director of supercomputer services at the Ohio Supercomputer Center and as XSEDE industry relations manager.

XSEDE Industry Skills Assessment Survey

The XSEDE Industry Relations team is gearing up to find out what skill gaps exist for new HPC recruits so that we can learn how to enhance existing, or create new certificate programmes for improved workforce development.

Industry Relations created a survey to better identify the training needs of industry with respect to computational modelling and high performance computing skills. The answers will help XSEDE share its training efforts with industry.

The survey should take less than 20 minutes of your time and individual responses will be kept in confidence.

To take the survey, click here.