HPC Ecosystem Creates Community

This writer has a calculation he can’t solve, so let it be a challenge to the programmers: how many hands directly touch a single supercomputer, and how many more are involved indirectly?

Take a moment to reflect on all the hours and effort that go into getting a high performance computer (HPC) ready to run its first job. It’s so much more than the people who build the data center and put the system together. 

It’s the people who mine the raw materials, work in the factories that assemble the components and move them along through the supply chain. It’s the vendors who create the chipsets, thermal controls, and networking equipment, along with the programmers who work on the software stack. It’s also the people who support the ecosystem: the investors, business leaders, sales and marketing teams, and even writers like me. We all play a role in the community, and it’s in service of one goal: creating and maintaining supercomputers that are used daily to push the limits of science, technology, discovery, and innovation.

No single organization can build an HPC system on its own. Even from an engineering perspective, these machines require deep expertise across many niche domains.

We need each other. In coming together to build the world’s most powerful computers, something else emerges as well: a community that brings us together to foster friendship, collaboration, and international cooperation.

Open Software & Standards Benefit Everyone

During my conversations at SC23, many people kept coming back to certain touchpoints for collaboration. Open standards and open source software form a bedrock for the HPC community because they benefit everyone and promote interoperability.

One of the most enduring examples is OpenMP, an application programming interface, or API, specification for parallel programming. Charles Archer, Chief Technology Officer at Cornelis Networks, calls OpenMP “a flag that the entire community can rally around,” citing an ever-increasing need for parallelization in the current artificial intelligence (AI)-era.

“Now OpenMP is an association of vendors that collaborates in committees, which focus, for instance, on accelerator features.”

This is the next chapter in a long story. OpenMP has been around since 1997, and its prominence has only grown. “It started with a small group of vendors,” explains Dr. Matthijs van Waveren, Marketing Coordinator at the OpenMP Architecture Review Board. “Now OpenMP is an association of vendors that collaborates in committees, which focus, for instance, on accelerator features.”

OpenMP certainly isn’t alone in this spirit of collaboration. The Ultra Ethernet Consortium was formed in July 2023 to bring the industry together to create an interoperable networking architecture for HPC and AI at scale. At the same time, the Innovative Optical and Wireless Global Forum (IOWN) is promoting collaboration among networking equipment vendors to lower power consumption, widen bandwidth, and improve interoperability.

Meanwhile, the Compute Express Link (CXL), is an open standard interface for interconnecting central processing units to other hardware components, such as random-access memory (RAM) and graphics processing units (GPUs). 

“A lot of time, memory is stranded and becomes wasted,” says Jeff Demain, Business Development Manager at Ayar Labs. “CXL, especially when combined with optical I/O [input/output], means that processors can more effectively share memory and optimize resource usage.”

Linux a Cornerstone of the HPC Community

No discussion of the HPC ecosystem is complete without talking about Linux, the open source operating system that’s at the kernel of pretty much every supercomputer. There are many flavors and distributions to choose from, but few would deny that Linux is a foundational element of HPC that benefits everyone.

This collaboration extends beyond the operating system (OS) itself. The Linux Foundation recently announced the formation of the High Performance Software Foundation (HPSF), an effort to support HPC open source projects more broadly. Their 10 initial open source technical projects range from HPC package managers to container systems to reusable scientific software libraries.

The OpenHPC Board collaborative added that its members are looking forward to collaborating on the HPSF as “OpenHPC provides a good base on which to install high-level software and run applications like what HPSF will deliver.”

Going back to Linux, at SC23 I had a great chat with Jonathan Wright, Infrastructure Team Lead at the AlmaLinux OS Foundation. His organization’s distribution was a direct community response to the death of CentOS. Today, it’s an essential part of the supercomputers at top research institutions like CERN and Fermilab.

“The Large Hadron Collider runs on AlmaLinux,” says Wright. “Our OS is a piece of what makes that possible. I didn’t think that I would ever have the opportunity to contribute, but now I’m a part of something larger than myself.”

HPC Creates an Inclusive Community

In HPC, there’s room for everyone, regardless of their background or location. At SC23 people from around the world came together to celebrate their love of supercomputing and connect with the extended HPC community.

From Japan to South Africa, from Korea to Saudi Arabia, from Mexico to Germany—and everywhere in between—the HPC community depends upon and promotes international cooperation. Vendors from at least 31 countries set up booths in the exhibit hall last year, and this year is trending towards even more global participation.

As vendors from diverse nations converge in our exhibit hall, the HPC community is also working to make a more inclusive environment for all. “In order to enable our next generation of scientists and researchers, we need to work towards reducing barriers and creating equal opportunities to leverage compute and AI resources,” says Wil Mayers, Technical Consultant at Alces Flight. “With the recent focus on Artificial Intelligence and computer learning, building a diverse community that represents all of our beliefs and values will become essential to developing and optimizing the technologies that will play a pivotal role in our future.” 

“There has been an upward trajectory in allyship that has come about in the last few years for women and those from underrepresented groups. Our community has taken notice of the need for diversity, and no longer considers the work done in this space as ‘niche’ or a background project.”

Alces Flight’s Chief Marketing Officer, Cristin Merrit, who also serves on the Women in HPC (WHPC) Executive Board, agrees. “There has been an upward trajectory in allyship that has come about in the last few years for women and those from underrepresented groups. Our community has taken notice of the need for diversity, and no longer considers the work done in this space as ‘niche’ or a background project. The continued trends for larger, more complex technology to solve global problems has helped guide us to this point, and now is a great opportunity to become involved in our field. We truly wish to welcome everyone to join in.”

Registration Opens July 10

Now, it’s your turn to get involved in the HPC community. The upcoming SC24 Conference in Atlanta will be the place to create new relationships, build partnerships, and collaborate with people from across the industry and around the world.

Registration opens in July. Hope to see you there! 

ABOUT THE AUTHOR

John Himes is the founder of Dynamic Tech Media, a content marketing agency that translates tech-speak into plain English for HPC, quantum, and other complex technology companies.