Large-scale experiments at CERN and the KM3NeT neutrino telescope in the Mediterranean Sea are generating far more—and more complex—measurement data than the current computing infrastructure can handle. NWO is providing 3.5 million for a major upgrade.
This was announced by NWO today in The Hague. In total, the research funding agency is awarding 30.5 million euros to ten projects aimed at making the existing large-scale infrastructure for scientific research more effective. A large portion of this funding is related to computing capacity.
This upgrade is particularly urgent in fundamental particle physics. The computer infrastructure that made the 2012 discovery of the Higgs boson possible through CERN measurements, for example, is now too limited for modern computational techniques such as machine learning and the volumes of data involved. As a result, particle physicists know that many details in the measurement data remain untapped—and with them, opportunities for new insights.
Under the title MERGE (Machine Learning E-Infrastructures for Results on the Global Exascale), Nikhef and the university computing consortium SURF will work on a new computing infrastructure for massive amounts of data. The project will fully leverage artificial intelligence and the hardware and software needed to support and streamline it.
David Groep, program director of computing for research at Nikhef, is pleased with the grant. “With MERGE, we can push the boundaries of computing capacity in all kinds of directions—for large-scale artificial intelligence in research, for more efficient and energy-efficient computing, and for the transition from small-scale analysis projects to production-scale operations worldwide.”
In the joint project with SURF, the Snellius supercomputer at Sciencepark Amsterdam will, among other things, be fed with significantly more data from the experiments at CERN for analysis using advanced artificial intelligence and machine learning. State-of-the-art graphics processors and accelerator cards will also be utilized. SURF will be responsible for distributing data streams and computing capacity.
PDP Group leader Groep also calls the planned upgrade a challenge. “Building something like MERGE is an interesting and exciting engineering task,” he says.
Groep, coordinator of the funded proposal on behalf of all Nikhef partners and SURF, emphasizes the absolute necessity of the upscaling. Already, the volume of data streams from projects at CERN and KM3NeT is many times greater than the computing capacity required to analyze them.
With the arrival of the high-intensity LHC accelerator and the completion of KM3NeT, we’re looking at exabytes of data—a billion gigabytes.
For Nikhef Director Jorgen D’Hondt, the MERGE grant aligns with his goal of making optimal use of the capabilities of artificial intelligence in data processing for particle physics. “This is the only way we can delve even deeper into unknown areas of physics,” he says enthusiastically. “Now and in the future.”
Earlier this year, Nikhef established a special AI task force to investigate and promote the effective use of artificial intelligence in particle and astroparticle physics.
The new MERGE infrastructure is explicitly intended to benefit Dutch science as a whole. In principle, projects other than LHC experiments and KM3NeT can also make use of it, according to the approved proposal.
