Re: RAID 1 for critical dbspaces...

On Thu, 31 Jul 2003 21:01:57 -0400, Ronald Cole wrote: > "Art S. Kagel" <kagel@bloomberg.net> writes: >> Since switching to RAID10 5 years ago NONE of these errors has recurred >> and believe me we've lost our share of drives, power supplies, and >> controllers over this period. > ... >> Sounds like it's past time for my NO RAID5 rant. I'll repost it >> shortly, look for it. > > Ok, I'm convinced to switch from RAID1 to RAID10. I have a PowerEdge > 1750 with hardware RAID 0, 1, and 5 on it's SCSI controller. How do I > configure it for RAID10? Or am I stuck with RAID1? > WELLL, you have several options. Switch to a controller that supports RAID10, replace the standalone disks with an array that supports RAID10 with multiple controller paths. There are several excellent SCSI/RAID controllers that support RAID10 (and one IB that supports RAID01 with hardware single drive recovery so it's functionally equivalent to RAID10 in every way that matters). However, ideally you should not mirror drives within a single channel controller. If the controller fails you are down until it can be replaced! There are a couple of dual channel controllers that also support RAID10 which is better but again you have to schedule downtime to replace the controller in order to restore the mirrors so I'd still recommend an intelligent array with built-in redundant controller paths and power supplies. Another option is to implement the RAID1 in hardware and stripe the mirrors using a software Volume Manager (VM) like Veritas or the Linux VM. You can also go all software but the performance will not be up to the hardware solution's level. We implement RAID10 in hardware and plaid (ie stripe multiple stripe sets) them using Veritas. Art S. Kagel