Research giant reaps rewards from Oracle


29 Jan 2004

While many organisations might drool over the prospect of evolving IT systems, others simply make it their business to adopt them as quickly as possible because new technology is inseparable from their business model. Such is the high-tech approach of CERN, a joint European organisation for nuclear research and the world’s largest particle physics centre. Dependency on cutting-edge technologies is, quite simply, fundamental to its existence.

With an agenda of cutting complexity at the same time as managing larger and more integrated pools of data, it is not surprising that CERN has struck up a long-standing relationship with Oracle, an IT vendor that sings from the same hymn sheet when it comes to reducing complexity while handling ever larger information stores.

“We see a lot of synergy with the direction in which Oracle is going. We agree with its philosophy,” says Jamie Shiers (pictured), CERN’s database group leader. “In the early Eighties an Oracle database was brought in to help with the construction of a big accelerator. Since then Oracle has spread across the whole spectrum of the lab’s activities, from internal e-business to more recent scientific research projects and for grid computing itself.”

While grid computing may have become a hot topic for most of us in the last couple of years, the boffins from CERN have long since recognised its possibilities. The concept of having clusters of server systems working together to create a single supercomputer is a perfect fit for the laboratory and its work. Founded in 1954, CERN is a collaborative centre where physicists from some 20 member states can avail of its research tools. As well as having 2,000 PCs in its Swiss laboratory and around 20 other sites around the world, CERN also collaborates with more than 500 institutes and carries out business with more than 20,000 suppliers.

“It’s about getting the most cost-effective computing power we can,” says Shiers. “There’s been an evolutionary approach. In the late Eighties and early Nineties we were still using mainframes. This was at a time when it hadn’t been demonstrated how to build a farm of high-end workstations and do bulk processing on them.”

The underlying challenge has been coping with an explosion in data, especially pertaining to its scientific work. “We have begun work on a new project that involves a thousand times more data than the project we started in 1989,” he explains.

To cope with the changing landscape, CERN took the decision to move to multiple connected desktops rather than a single mainframe. In CERN’s research, results of individual particle interaction are kept separate from any other with the data divided out across a vast farm of PCs.

With floating point computing used for storing and calculating numbers, CERN has placed demands on its suppliers that have helped developed closer working relationships when it comes to product development. In the case of Oracle, product such as its new 10g database software, where the ‘g’ stands for ‘grid’, has taken onboard the input of CERN’s technical team to deliver performance and storage improvements.

“Databases traditionally come much more from the commercial world where you’re normally storing data such as dollars or cents, which they do efficiently,” explains Shiers, “but the way they stored floating point numbers, like Pi and other calculations, was not efficient. We’re using 9i and talking to Oracle about moving to 10. It has added some valuable new features.”

There is also a feature that enables the user to partition part of a database, unplug it and reconnect it to any other Oracle database anywhere in the world. For CERN this is a valuable collaborative tool for sharing data in other labs and universities using the same data.

Outside of the particle physics research, Oracle also sponsors a grid test bed in CERN’s openlab facility and has also become important to the organisation in its business systems. It delivers critical information for budget management, cash flow forecasting, project management, and cost-to-completion analysis to serve more than 1,000 users and process approximately 1,000 sophisticated analytical queries daily.

Picture courtesy of CERN.