Bringing Powerful HPC Software to Small- and Medium-Sized Manufacturers 

The Department of Energy's Office of Advanced Scientific Computing Research (ASCR) has joined the ranks of agencies, institutions and individuals determined to help small- to medium-sized manufacturers (SMMs) benefit from the use of advanced digital manufacturing technology.

ASCR is a welcome addition, bringing both a treasure trove of HPC software as well as grant money to the table.

The mechanism for engaging with the SMMs is a Small Business Innovative Research (SBIR) topic that was issued by the Department of Energy in mid-July to solicit grant applications. The grant application period is now underway.

"The SBIR strategy is to write topics that resonate with the members of the manufacturing community," says Dr. Daniel A. Hitchcock, acting associate director for ASCR. "The current topic is titled 'Increasing Adoption of HPC Modeling and Simulation in the Advanced Manufacturing and Engineering Industries.' Our goal is to make software and solutions available to companies that have yet to adopt HPC into their product development lifecycle."

Slow Adoption

Hitchcock notes that despite the great potential of modeling and simulation (M&S) in aiding the understanding of a variety of important engineering and manufacturing challenges, HPC has been underutilized in the manufacturing community. Roadblocks include the complexity of the applications, the need for considerable in-house expertise, and a perception among the SMMs that HPC hardware and software comes with a hefty price tag.

Over the past ten years, ASCR has invested millions of dollars in the development of HPC modeling and simulation software and tools. For example, the Portable Extensible Toolkit for Scientific Computation (PETSc) is a suite of data structures and routines for the scalable (parallel) solution of scientific applications modeled by partial differential equations. It's a powerful piece of software that can be applied across a number of domains. PETSc is also very hard to use — a high degree of computational sophistication and expertise is mandatory. It's also award-winning software — in 2009, PETSc Release 3.0 was named a R&D Award winner by R&D magazine. The prestigious award celebrates the top high technology products of the year.

Another available ASCR tool is ScaLAPACK (Scalable Linear Algebra Package), designed for heterogeneous computing. This is another highly useful but complex set of software routines that also require an expert at the helm.

So it appears that the biggest stumbling block to HPC adoption among the SMMs is the difficulty of using the advanced modeling and simulation software, a problem compounded by a lack of much needed support.

In Hitcock's opinion, "It's a problem we face at the national labs as well. We spend a lot of time training people to learn to use these codes — the software is difficult and requires a major investment in time and money before the individual can hit the ground running. Parallelism adds another layer of complexity — but you can't do without it. Parallelism is essential to making full use of HPC systems, especially as machines ranging from workstations to supercomputers become increasingly more powerful with the addition of multiple cores, massive storage and high speed interconnects. Because parallelism is still not taught in many universities, you have to learn it on the job — another limiting factor. This SBIR topics from ASCR could mitigate these challenges faced by many industries."

Bridging the Gap

The SBIR strategy is to write topics that bridge the gap between the SMMs that want to use advanced digital manufacturing tools and organizations like the Department of Energy and its leadership laboratories, and educational institutions that are willing to share their expertise but can't provide support.

The SBIR topic addresses this challenge by requesting grant applications in three specific subtopics:

1. Turnkey HPC solutions for manufacturing and engineering — Here ASCR is looking for HPC applications that can used both in the advanced manufacturing supply chain, and by the less skilled, more general user. In the latter case, applications must be easier to learn, test and integrate into the product development cycle by users who have some computational experience but are not necessarily experts. These applications should port to HPC platforms that are more reasonably priced when compared to current high-end systems.

2. HPC support tools and services — The HPC community is rich in tools and services that have been developed over the years. These debuggers, profilers, workflow engines, libraries, data management and other HPC tools are very powerful. However, they also take a considerable time to learn and use, making them essentially unavailable to many manufacturers. ASCR is seeking grant applications from ISVs and others to help make these tools easier to use for the experienced but not expert user. Potential solutions include enhanced or simplified user interfaces, a more intuitive interaction with the user, and consolidation of related tools into a common environment and framework.

3. Hardening of an R&D code for industry use — ASCR has invested millions of dollars in the development of modeling and simulation software, solvers and tools. Many are open source and designed for use by highly trained experts. Their level of complexity, while essential for knotty scientific problems, may not be needed for industrial applications. Grants will be awarded to companies that can take portions of the codes developed through SciDAC (Scientific Discovery through Advanced Computing) and other ASCR programs and "shrink wrap" them into tools that can be handled by users with a lower level of expertise. This might be accomplished by designing intuitive GUIs, simplifying user input, lowering the complexity of the code by stripping out components, and providing user support tools and services.

The grants include $150,000 for a Phase 1 feasibility study. Another $1 million is allocated for Phase 2 prototype development over a two-year period.

Hitchcock says ISVs are key to the success of the program. They provide the bridge between the SMMs and ASCR's software resources. "Funded by the grant program, their role is to write the interim software that makes the existing HPC software usable and affordable for the manufacturers. We're taking a multi-year approach. Initially we are addressing the larger SMMs — those that have quite a bit of HPC savvy but have not really leveraged their knowledge. Then we will write more topics as we address the needs of smaller manufacturers that can also benefit from advanced digital manufacturing capabilities."

Growing the Field

Hitchcock points out that although larger manufacturers are directly involved at this point, the software developed for SMMs by the ISVs may also be of great interest to the P&Gs, GMs and Boeings of the world. There's also the matter of synergy. By developing an HPC ecosystem among the SMM "missing middle," chances are that a number of the smaller firms will in time evolve into larger organizations. As they grow, so do their HPC requirements.

"This program is helping us fill the pipeline with organizations that will not only make good use of HPC, but get to a point where they turn to the DOE labs and their massive supercomputers for help processing highly complex, data intensive jobs," Hitchcock says. "In the process we help invigorate the overall manufacturing sector, which creates jobs and allows the U.S. to be more competitive in the global marketplace."

For more information:

• On DOE's SBIR program go to http://science.energy.gov/sbir/.

• Email Rich Carlson, program manager, at richard.carlson@science.doe.gov.

• On the Office of Science, the single largest supporter of basic research in the physical sciences in the United States, visit http://science.energy.gov.