500PB of scientific data is now managed by STFC Scientific Computing, which says rising output from major research facilities is pushing scientists toward near-real-time analysis and AI-driven workflows.
Ada, STFC’s remote analysis platform, delivers Data Analysis as a Service to thousands of researchers, typically supporting 50 to 100 concurrent users without requiring specialist computing training.
400PB on tape and 190PB on disk sit behind a network exceeding 30 terabits per second, while inbound data to the Rutherford Appleton Laboratory averages 600Gbps a month, including about 200Gbps from CERN’s LHC.
AI for Science is being built through the Ada Lovelace Centre, which is preparing experiment data for model training and backing tools such as the MACE-MP materials model to speed simulations across chemistry and materials research.
JASMIN extends that infrastructure into climate and Earth observation research, serving more than 2,500 users with 90PB of disk, 100PB of tape and a recent £3 million upgrade from the Natural Environment Research Council.
With scientific data streams growing exponentially, is centralized supercomputing sustainable, or will localized edge computing at the instrument level become the new standard?
As AI powers autonomous labs to process exabytes of data, could algorithmic bias silently compromise the foundations of future scientific breakthroughs?