Kyoto University Loses 77 Terabytes of Research Data After Supercomputer Backup Error
A routine backup procedure meant to safeguard data of researchers at Kyoto University in Japan went awry and deleted 77 terabytes of data, Gizmodo reported. The incident occurred between December 14 and 16, first came to light on the 16th, and affected as many as 14 research groups at the university.
Supercomputers are the ultimate computing devices available to researchers as they try to answer complex questions on a range of topics from molecular modeling to oil exploration, climate change models to quantum mechanics, to name a few. Capable of making hundred quadrillion operations a second, these computers are not only expensive to build but also to operate, costing hundreds of dollars for every hour of operation.
According to Bleeping Computer that originally reported the mishap, the university uses Cray supercomputers with the top system employing 122,400 computing cores. The memory on the system though is limited to approximately 197 terabytes and therefore, an Exascaler data storage system is used, which can transfer 150 GB of data per second and store up to 24 petabytes of information.
Although the details are not available yet, an error during the routine backup process ended up deleting approximately 34 million files belonging to 14 different research groups that accessed the supercomputer. When the error was discovered, the backup process was halted, an official university statement said. Initial estimates suggested that nearly 100 TB of data may have been lost, however, further investigation revealed that damage was considerably lesser.
For most of the research groups, some data could be recovered through a backup but four of them have lost their research data forever. As of now, the university has scrapped the backup process altogether and will reintroduce it in 2022 with plans of using incremental backups - which covers files that have changed since the last backup, as well as complete backups, Bleeping Computer reported.
Details about the research groups or the nature of research impacted by this mishap have not been revealed by the university.
The internet-era adage of 'always backup your data' holds good for supercomputers too. Maybe we need to add, 'Do not trust tech support to do it' as well.