Was speaking with a friend the other night, they’re having an internal debateÃ‚Â (to IT that is) about what kind of disks to have.Ã‚Â The old school IT view is RAID, replicated storage so if one disk fails, the blade is still in service.Ã‚Â The main views in this respect appeared to be:
* RAID 1 mirrored set (two disks)Ã‚Â -Ã‚Â Ã‚Â If the disk fails, life continues, obtain the downtime replace the disk and we’re back in service.
* Single IDE/SCSI disk – Keep it simple, one disk, if it fails, replace the disk, rebuilt it, there’s no need for RAID particularly if it’s a calculation engine with no historical data.
* No local disks – san boot it, have everything on emc (or the equivalent) that way you save the power/heat generated by local disks and allow a more fluid type blade infrastructure.
The debate from what I gathered as we discussed over dinner, is set to continue, SAN boot? Can our switch handle it?Ã‚Â How do you build on a network attached disk? What about drivers?
Also,the thought that a server platform would be ‘production ready’ with an option to rebuild like a desktop, if a disk fails or software installation gets corrupt is not the typical way of operating in the ‘olden days’ infrastructure world, where disks are plentiful and servers last forever. Bring back my compaq proliant 2005 with it’s local 7200rpm disks, with windows nt, when life was simple.