Into the future in partnership


Innovation needs concurrence – one of the important basic idea of the new procurement method innovation partnership. Photo: Nejk Soklic/Unsplash

The next system shall contain state-of-the-art technology, components that are not (yet) developed and that offer research new possibilities: For the procurement of its next supercomputer, the Leibniz Supercomputing Centre (LRZ) is relying on a new procedure, the so called “Innovationspartnerschaft” or partnership for innovation. The architecture of the new supercomputer at the LRZ will be developed in a co-design approach together with technology companies until 2024, with prototypes being built or existing components adapted to specific needs. "It has been clear since the 2010s that the next generation of HPC systems, exascale computers, will have to be developed in a co-design approach, meaning that hardware and software will be optimized in coordination for the desired application. In the academic community, many U.S. National Labs already rely on co-design of hardware and software," explains Prof. Dr. Dieter Kranzlmüller, director of the LRZ. "We wanted to improve the procurement of our systems, and the partnership for innovation lent itself to that." The funders, the German Federal Ministry of Education and Research (BMBF) and the Bavarian State Ministry of Science and the Arts, also supported this new approach. Hence, the next national supercomputer for Garching can be planned as the first in Europe within the framework of a partnership for innovation. "Suited to our needs," Kranzlmüller expects, "we can research and evaluate innovative computing approaches together with manufacturers and finally implement them in the new exascale system."

Tool for modernisation and business development at the same time

Partnerships for innovation have been introduced as a tool in Germany in 2016. The federal and state governments support this procurement method because it supports the development of new offerings and innovative companies, and even helps modernize administration and public institutions, according to the Federal Ministry of Economics. "The public sector spends around 500 billion euros per year in Germany," notes Matthias Berg, head of the “Competence Centre Innovative Procurement”, KOINNO for short. He has already accompanied several such partnerships as a consultant. "If only one percent flows into new products and services in the process, that's an enormous economic boost." To be sure, the development of the exascale computer is less about modernization and economic development than it is about developing promising computer technologies. Like other innovation partnerships, the LRZ will proceed in several phases. Around ten specialists from various fields are involved in the procurement process, which began with a competitive bidding process at the start of 2022. In several rounds and until mid-April, hardware companies competed according to the LRZ's specifications with concepts for the next high-performance computing or HPC system. Currently, the bidding companies are invited to submit official bids with detailed co-design steps to the LRZ. In July, the partners for the next phase will be decided on and contracts will be in place. The R&D work on the new system will begin. By 2024, LRZ will decide which company will build and support the final system, the first Exascale-generation supercomputer at the LRZ.

"The partnership for innovation enables us to develop and optimize technologies for the next system with several suppliers at the same time," says Herbert Huber, citing advantages. Huber, who holds a doctorate in physics, has headed the High-Performance Systems department at LRZ for ten years and has helped plan and procure several supercomputers. "This allows us to get to know the companies and how they work well before the supercomputer is delivered." LRZ's innovation partnership incorporates experience from Future Computing and with the Bavarian Energy, Architecture and Software Testbed (BEAST) test environment. The Big Data & Artificial Intelligence Team (BDAI) and the Computational X Support (CXS) group, which supports researchers in optimizing algorithms and implementing them in HPC systems, are also collaborating on proposals and specifications. Hopes are high that the interaction of the supercomputer's components can be tested and improved in advance. They also hope that teething problems will decrease as the system becomes operational. "Once the initial technical steps and contracts have been clarified," Huber plans, "we will therefore also involve users in the development work."

Effort, discussions and risk

Partnerships for innovation are more costly than other procurement procedures. "They are worthwhile for products with high investment requirements and for new developments," observes Berg. "However, there is usually only a vague idea of the technology or service to be developed, and contracts are concluded for research and development work and prototype construction." These touch on copyrights, patent rights and also marketing rights and are therefore drawn up in a more individual and detailed manner. There are also many precautions to take. Among other things, the LRZ must ensure that all companies can work with the same information and that no one learns about the technical innovations of the others. The prototype processors, accelerators and other components are therefore installed in separate and locked racks. And the team is sworn to secrecy; who contributes what and how to the next supercomputer is currently not allowed to go public.

Innovation expert Berg also recommends not underestimating the time required. "For the individual development steps, the design specifications and, above all, the test criteria should be carefully defined and worked out," he says. "As many people as possible should be involved in the discussions so that different perspectives and needs are taken into account." Even if several companies are working on a problem at the same time in the first rounds and are rewarded for it, innovation partnerships definitely help save money, too. Experience shows that costs in the IT sector can be reduced by up to 30 percent because hardware or software is developed more specifically to meet real needs. Above all, however, in the case of the new LRZ supercomputer, the best possible exascale system for research and science is to be created through technological competition - and hopefully new components will also be developed along the way that will help to better design IT processes in industry. In the innovation partnership, participating companies run the risk that their solutions will not ultimately be selected and ordered. On the other hand, they can market the products developed during the collaboration themselves and adapt them for further applications. "It's quite possible that ideas will fail, but these experiences are valuable for everyone involved. Failure is part of innovation, and the public sector has a hard time with it because it has to account for its spending," says Berg confidently: "Although the process is complex and the risks are high, I've yet to see an innovation partnership fail."  (vs)