I've done fairly extensive research on the 10K SATA 150MB Raptors vs 10K or 15K SCSI for database purposes.. Basically the Raptors perform much better than any other IDE/SATA drive out there, but the SCSI drives perform much better than the Raptors. So it's like the Raptors get you about 40% of the way to SCSI but you will still suffer. Especially with a lot of small reads. Now if you go to 15K SCSI there is no comparison.
I use RAID5, but I am going to use RAID10 on any future setups. It is also is smart to have a secondary drive outside the RAID array for backups or log files. By using a secondary drive for backup creation, I cut my backup time by about half (and cut it to 1/4 by using incremental backups without compression instead of compressed files). So it went from 4 hours to less than 1 hour. This is useful even if you store your backups off-server (as you should) since the backup creation still usually occurs locally before it is FTP'd off somewhere.
Vissa,
In my new server I have a extra 300 Gig HD that is outside the RAID system for this purpose. It has not been racked yet (hopefully Monday) and setup but I wanted to ask you since you seem to have a similar setup, how often you backup outside of the server and if you do it in that data center or another one, or local?
I backup to the separate HD nightly (daily, weekly and monthly retained), then every few days or once week off-site to a server at another datacenter. I could have used the datacenter's backup solution but this is free for me and I have the bandwidth to spare (backing up between the 2 servers from different datacenters). Just make sure you use SSH2 for transfers off-site. So if there is an earthquake/fire/etc at a datacenter I'm still ok.
I like the idea of backing up to a separate server elsewhere.
If there was a catastrophic failure at the primary DC or with the server(s) there, how long would it take to get back online at the secondary DC? Is everything setup and ready to roll as soon as you change IP's ?
What about the server spec's at DC #2? Are you using a lower end server or do you really need to have a dupe of the primary? That would be quite expensive for me but having a lower end one (maybe a single Opteron 270 with SATA RAID 1) could be feasible.
This would be a cool topic for a new thread in the BB section: What type of redundancy / fail-over do you have for your big board.
I would go with the 15,000 RPM drives for sure, but the u320 SCSI setup will be fine. SAS is new and in the expensive stages. I don't know how much of an increase it actually gives right now. I went for it more out of future upgrade compatibility than anything else.