The CERN Data Centre processes a staggering 1000 terabytes of data every day, which is the equivalent of around 210,000 DVDs. And, at peak times, the Worldwide Large Hadron Collider (LHC) Computing Grid may transfer as much as 10 gigabytes of data from its servers every second. However, CERN is not the only research organization dealing with such 'big data'. Ever-cheaper sensing and data-storage technologies are driving an explosion in data across many fields of science, as well as for businesses of all sizes.
Yesterday, CERN openlab invited research institutions and leading IT companies to participate in a workshop focused on the latest advances in data analytics. "Data analytics brings the promise of further increasing the efficiency of LHC operations, accelerating physics analysis, and preparing us for even more ambitious initiatives," says Alberto Di Meglio of CERN openlab. "The next CERN openlab phase will lead us to the start of the next planned shutdown of the LHC (LS2) in 2018. This is an excellent opportunity for us to investigate new technical challenges with commercial partners and other European research laboratories."
Presentations were given at the event by representatives of the invited IT companies. Africa Perianez of the German Meteorological Service spoke about how big data is driving advancements in weather prediction, while at the same time posing analytical challenges. Meanwhile, Nenad Buncic of the Swiss Federal Institute of Technology in Lausanne (EPFL) gave a presentation on the Human Brain Project, an ambitious attempt to improve understanding of the brain by building a completely new type of computing infrastructure. Whereas at CERN the vast majority of scientific data is generated centrally by the LHC experiments, the data for the Human Brain Project is produced by a host of smaller, distributed laboratories. "A lack of data sharing between labs is probably the biggest challenge in neuroscience," says Buncic. "We are incentivizing labs to share data… but we need to make sure that data conforms to standards and that researchers speak a common vocabulary."
Salim Ansari from the European Space Research and Technology Centre also gave a presentation on the vast amounts of data being generated by the European Space Agency's fleet of Earth-observing and space-science satellites. He discussed how the operational efficiency of satellites can be improved by analyzing orbital data and explained how the concept of 'predictive analytics' is changing how astrophysical data is handled. "Analytics can play a major role in simplifying scientific research," says Ansari. "Discovery tools can be automated, thus freeing the scientist from having to carry out much of the search for knowledge."
Finally, CERN's Johannes Gutleber gave a brief presentation on long-term studies into a potential successor to the LHC. While any such machine wouldn't begin operations for at least another 20 years, it is important, says Gutleber, to think now about the computing infrastructures that will be required to analyze the huge data output. "We need to think about how the computing infrastructures are going to evolve. We need to build infrastructures that are going to survive the trends and fashions of computing today, while being flexible enough to incorporate future advancements in technology."