Cleversafe has improved the hardware it sells to distribute slices of data around a network, reckoning this is much cheaper at the PB scale than a RAID and replication combination. RAID 5 will protect a drive failure. RAID 6 will proptect against a double drive failure. So if a second drive fails while a rebuild is recovering …
Sounds an IDA storage to me
It might be an <Pedantic IT architect> IDA: Information Dispersal Algorithms. </Pedantic IT architect>
Worth to mention that there is a company called B-Virtual that successfully implemented this smart stuff long time ago and it sells it, among others, to Cloud providers.
"The storage overhead is said to be much less than with a generic RAID plus replication arrangement."
RAID 1: 50% of your storage space is usable
Replication + RAID 5 (the most efficient that's actually useful): <50% of your space is usable (your entire second DC is not usable as extra space, plus the drive you lose for RAID-5)
"The storage overhead of RAID 1 is said to be much less than with a generic RAID plus replication arrangement."
are they really that inefficient? or are they trying to imply that if you use their system you don't need to bother with replicating it because when you lose your primary DC it'll keep working by magic?
Or perhaps they expect you to run this as a single storage unit spread across multiple datacentres so a single datacentre doesn't contain more than a small portion of the drives? yes, let's have most of our data stored on the other side of a WAN link...