we need modern, efficient and resilient file systems (in the mould of ZFS) and we need to get rid of RAID arrays.
latest #8
Also, better error recovery at a more granular level then whole disks. RAID6 is a band-aid.
A file system that handles data redundancy at the block level, with forward error correction and checksums.
And then block replication / error correction across disks to handle disk failures.
And across bus / controllers upstream of that. Like GoogleFS but applicable to normal workloads.
Journaling and RAID was fine in the '90s with 10-40 GB disks and rebuild time of 40 minutes.
Can't believe the data protection systems I have to work with now are the same as in 1999
can't believe the alphabet I have to work with now is the same as it was in 1599
back to top