Old File Systems

/s/

• Linux global scratch file system
• files are not permanent but will be deleted periodically (see cleanup policy below)

Data file systems /d/ (being phased out)

• Linux data file systems on RAID systems on dedicated file servers
• Large volumes able to handle experiment - and simulation data
• Fileservers are bought by experiment collaborations, administration is done by the LinuxGroup, volumes are typically accessible as /d/groupnamesomenumber
• No automatic backup of the data is organised. Of course, you can archive data on the tape library manually.
• You share a volume with up to 100 users of your group. Parallel and concurrent read/write acesss will significantly slow down the I/O performance. Optimise your analysis to avoid parallel I/O with a high number of batch jobs writing on the same volume. Your colleagues will not be happy, if you bring down the I/O speed on your groups file server by writing to it with 500 batch jobs in parallel ...