Four years after passing the U.S. National Quantum Initiative Act and decades after early quantum development and commercialization efforts started – think D-Wave Systems and IBM, for example – the U.S. quantum landscape has become a roiling cauldron of diverse activity. It’s perhaps too easy to forget that the U.S. is hardly alone in catching the quantum bug. Europe has also jumped into the fray, as have China, Japan, Canada, Australia and many others. No one wants to miss out on what could globally become a transformational technology.
One area that Europe is tackling sooner than the U.S. is work to fully integrate quantum computing with traditional HPC infrastructure. While there’s an emerging consensus worldwide that quantum computing is likely to become just another accelerator in the heterogeneous advanced computing architecture, Europe is taking deliberate steps now to make this a reality and the Quantum Integration Center (QIC) at the Leibniz Supercomputing Center (LRZ) is an illustrative example.
Now turning two years old, the Leibniz QIC has two major objectives. It’s meant to be a user facility providing access to a variety of quantum hardware and software and assist in their development. Nothing new there, and it’s still early on in standing up quantum systems. The second mission goal is to integrate quantum computing with Leibniz’ traditional computing infrastructure, which includes new AI technologies such as a Cerebras system.
The broad idea is that in the future, Leibniz Supercomputing Center users may submit jobs and not necessarily know which of the underlying hardware options are doing the number crunching. Quantum will be another accelerator in a mix of accelerators ready for work. Creating the blended infrastructure to do that efficiently is at the core of the Leibniz QIC’s mandate.
“We are responsible for what we’re calling the Munich Quantum Software stack – that’s to be able to develop needed algorithms and software tools all the way through to running and managing applications on quantum resources and incorporating HPC. The HPC-QC integration is a big part of this. Also, we’ll develop this capability in a qubit-modality-agnostic way,” said Laura Schulz, head of Leibniz QIC, who was part of the team that wrote the strategic plan for the QIC.
“At the end of the day, our users should be able to utilize this technology with the simplest, cleanest path available. Some users will care about what system they’re actually on, and will want to be able to fine-tune the pulses on those quantum systems. Then you’ve got the other spectrum, users, like many HPC users, that are not going to care as much about what they’re computing on; they’re going to care more about getting the performance.”
Ambitious goals. Schulz recently briefed HPCwire on Leibniz QIC plans and progress. A key early milestone, said Schulz, is demonstrating a working HPC-QC stack.
“We’ve got an early quantum system (5-qubit) and we have an HPC test center [that comprise] a great testbed, literally sitting in the same room. If you came into the room, you see them literally next to each other. The first milestone is making sure that these systems are connected and that we can send jobs through the HPC to the QPU. It comes back to work on software development. We want to have these systems in place and to have the software to enable interaction – not two different software stacks running independently, but a single source software. Then we’ll get progressively better as we get other systems in,” she said.
Though these are still early days, the Leibniz QIC has been growing rapidly. On the hardware side, it currently has a 5-qubit superconducting processor from IQM, simulators from Atos (QLM) and Intel (IQS) and will add more QPUs and types. For example, there’s a 20-qubit system coming as part of the Q-Exa Project. At the moment, Leibniz QIC is focused on superconducting-based quantum processors but the broad goal is to avoid being locked into single qubit technology.
“We’re waiting on the neutral atoms; those are a little bit further down the timeline. For us, right now, it’s superconducting because it offers great opportunities for scaling. Each of these systems has its own flavor and benefits. Superconducting is great for scalability, but it’s not as stable. Ion trap has more stability, but you can’t quite scale it as much.
“We have these different systems that we’re building up and we are going on the postulate we’re going to have multiple types of QPUs in the ecosystem; there’s not going to be one winner, right, and the technology is too new to bank on any one particular [approach]. But by having a suite of different types of modalities around, we’ll be able to experiment,” said Schulz, who was selected this year as an HPCwire Person to Watch.
It’s worth noting the wide range of the Leibniz QIC’s constituency. It is part of one of six European supercomputing centers involved in Europe’s quantum computing development effort. It is also part of the Munich Quantum Valley (MQV). Here’s how MQV describes itself:
“As a hub between research, industry, funders, and the public, Munich Quantum Valley (MQV) is the crystallization point for the development of the full spectrum of quantum technologies. It promotes an efficient knowledge transfer from research to industry, establishes a network with international reach and provides tailor-made education and training opportunities in the fields of quantum science and technology.
“Harnessing three of the most promising technology platforms – superconducting, neutral-atom, and trapped-ion qubit systems – Munich Quantum Valley will develop and operate competitive quantum computers in Bavaria. In a unique holistic approach researchers develop all layers, from hard- and software up to applications.
“The Munich Quantum Valley collaboration unites research capacities and technology transfer power of three major universities and key research organizations: the Bavarian Academy of Sciences and Humanities (BAdW), the Fraunhofer-Gesellschaft (FhG), the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), the German Aerospace Center (DLR), the Ludwig- Maximilians-Universität München (LMU), the Max Planck Society (MPG), and the Technical University of Munich (TUM). Their joint work will advance quantum technologies at all levels for future use in science, research and industrial applications.”
Think Silicon Valley focused on quantum. Perhaps more than in the U.S., the interplay between industry and government-funded programs is fundamental. For example, the MQV interchange with the Leibniz QIC is extensive said Schulz.
“I haven’t paid as much attention to the American situation as much as I should. What impresses me about what I see happening in Europe is this early dedication to HPC-QC integration. We know that quantum is going to have to be trusted, and have to be fortified, and it’s going to come in to the supercomputing realm. I mean, quantum is high performance computing, right. It’s going to end up as another accelerator capability.
“The other thing that I’ve noticed is the partnership with industry. And while there is some of the early hype, some overly ambitious promises and all, but what I’m seeing, trend-wise, is the conversation is more tempered. We realize that there’s a lot of possibilities here, but also realize we’ve got several steps to go to get to that potential promise. The companies that we’ve been interacting with have that mentality, they understand that the possibility is there, they’re doing these proof of concept projects,” said Schulz.
Over time, of course, the market will determine which development approaches win. China has embarked on an aggressive centralized plan. The U.S. has a blend of DOE-funded National QIS Research Centers and a vigorous separate commercial quantum development community. Europe has a Quantum Technologies Flagship program and the European High-Performance Computing Joint Undertaking (EuroHPC JU) which named LRZ as a quantum site.
At ground level, Schulz is busily ramping the Leibniz QIC. Staffing has been a challenge. Headcount is currently ~24 and headed north of 40 by year-end, hopes Schulz who is actively hiring. It’s a multi-discipline group, with its share of physicists, but software workers currently comprise around 50 percent of the team. There’s also many external collaborations within the MQV.
Said Schulz, “My team is kind of this microcosm of the community. I’ve got computer scientists, software developers, electrical engineers, and quantum physicists, experimental and theoretical, I have this really nice little community that represents this bigger picture. What’s funny is some of the issues that we face as a team. We were just doing all of our annual reviews, and the HPC people were saying, ‘I’m good on the quantum more or less, but need to know a lot more about how this works.’ The quantum people were like, ‘I really need to understand HPC more, for example how does the scheduling work?’ We’re cross-training each other within our own team, to ensure that everybody has a baseline to understand how this comes together smartly.”
Mixing quantum computing and traditional HPC in the same facility has also prompted new challenges.
Schulz said, “With HPC, and energy efficiency, the whole infrastructure is already complex and has evolved on its own. But now we’ve got these cryostats that we’re taking care of and we’re having to change out the nitrogen on a particular schedule. We’re having to learn how to calibrate these things and how to maintain the calibration. We’re having to learn a whole new set of operational programs. We have to worry about all these other external factors – humidity, temperature, electromagnetic radiation. This is a new instrument that we have for the compute and we have to figure out how to bring into an HPC center.
“Some of this technology is coming straight out of the physics labs, and we’re going to startup companies for some of the early pieces of this. We’ve got to try to help them understand what it’s going to take in their evolution and their form factors and their stability to be to be able to leave the system alone and have it function at the same level of care taking as is an HPC system.”
During this developmental stage, creating a hybrid quantum-classical HPC infrastructure at Leibniz QIC is an all-hands-on-deck enterprise. That’s not practical long-term, said Schulz, “We want to get to the point where we don’t have to have trained and skilled experimental physicists on staff 24/7. That’s a little extreme to have dedicated experimental physicists taking care of these systems. We see ourselves, at this point, helping the maturation of these technologies where they can exist in an HPC center. Hopefully, as these systems become more commercially viable, we’re hoping to help them exist in a commercial market space.”
Lately there has been buzz around chasing quantum advantage – the notion of using quantum technology in a commercial application. Schulz urges both patience and a change in thinking.
“When I hear quantum advantage, I know it’s usually used as metric for beating a classical computer in a particular application. I want to challenge that and suggest that there are other ways we should be thinking about quantum advantage. I think that for particular types of algorithms, for particular applications, or parts of an application, quantum is going to be fantastic or has the potential to be fantastic. However, that’s when we have a real-world application, an assembly of algorithms involved, and all of that has to work together. Quantum may be able to take part of that load off of this overall application.
“I’m looking at it from the HPC-QC integration and how all of this works together. So, I’m thinking about what is the HPC-QC advantage? What does that mean? I mentioned things like energy. So, energy may end up being an advantage for quantum over HPC. There’s other parameters that we should be thinking about. I know that everybody’s been shooting for that (quantum advantage). I think that that’s going to be a bit farther out. Let’s be honest, you know, there’s a lot of friction points that we have to sort out along the way.”
It will be interesting to monitor how QC-HPC integration efforts proceed at Leibniz QIC. The notion of a single stack able to manage multiple qubit modalities and systems and traditional HPC resources together, seamlessly, is enticing. Stay tuned.