The Large Hadron Collider (LHC) wakes from a three-year shutdown next month when beam tests begin ahead of experimental work. During the downtime, upgrades were made to the computer centres that handle the vast amounts of data produced when particles smash into each other close to the speed of light. But, despite the LHC being the most expensive scientific instrument built, the information it collects is still archived on magnetic tape – a technology that has barely changed since 1952.
Alberto Pace at the CERN particle physics laboratory near Geneva, Switzerland, which hosts the LHC, says that when it comes to storage “we are looking for cheap”, and tape is more cost-effective than hard discs or flash memory.
This isn’t to say that CERN avoids using hard discs. In fact, the Large Hadron Collider data centre contains around 90,000 of them, with a combined capacity of 487 petabytes – enough to store 974,000 years of MP3 music. But they are only used for short-term analysis: the content is stored in duplicate on hard discs for no more than a few months and is then deleted to make way for new experiments.
Advertisement
Before this happens, the content is copied onto 380 petabytes of tape storage for permanent backup, and duplicated in data centres around the world.
Tape carries many advantages, says Andy Walls at IBM, which invented tape storage in 1952 and which supplies storage hardware to CERN. If stored carefully, tape can last for decades. This makes it more reliable than hard discs, 30 of which fail weekly at the LHC. It also consumes no power while sitting idle.
Data stored on tape can later be moved back onto hard discs for fresh analysis. “Sometimes [the physicists] see something interesting in their analysis,” says Pace. “So they want to rewrite the software and recall the data from tape.”
There are still costs associated with tape storage at this scale. CERN has 50,000 tape cartridges for permanent backups, and requires robots to physically insert, remove and file tape cartridges as needed. Hard discs or flash storage would remove the need for these overheads – but would still not come close to beating tape on cost.
Walls says 1990 was the first time someone told him tape storage was dead. Nonetheless, he says it remains the best option for large operations.
“If you have to deal with tens of petabytes, exabytes of data, and you need to do so inexpensively, economically, tape is a way to go – especially if what you’re doing is kind of a sequential throughput where you’re bringing in lots of data at one time,” says Walls. “I don’t see anything on the horizon that’s going to overtake or eliminate it.”
