Problem solve Get help with specific problems with your technologies, process and projects.

Overhead for RAID 0+1 is too high ... should I use RAID-5?

I am in the process of migrating data from DASD to an enterprise-based SAN. This is one of several activities in a disaster recovery solution. A by-product of this is that I'll be left with lots of unused storage space from the legacy devices. I have a plan for these, but need some advice!

I plan to use two of the DASD (Sun 3310, 12 * 73 GB disks) to store replicated images of the boot disks from production machines. The idea here being that these can be used on the DR site to jumpstart a machine and have it quickly rebuild as a production machine, identical to how it was on the primary site before any disaster.

In terms of the storage devices, what is the best way to utilize the two 3310s? I have 24 disks, totaling around 1.7 TB of raw data. RAID 0+1 seems an option, but the overhead of this just seems obscene. How best can I utilize these 24 disks?

In terms of I/O performance, writing is not a primary consideration, but reading of data needs to be at a decent enough rate to not cause a bottleneck when jumpstarting clients. I also need to consider the fact that the units I have inherited are JBOD and do not have a hardware RAID controller.

I would definitely use RAID-5 in this situation. It has good read performance, and makes much more efficient use of the disks. I would not use them unprotected (they WILL fail!) and certainly would not waste half my space on 0+1! The 3310 can have dual SCSI busses, so RAID-1 might seem logical, but RAID-5 is the way to go. Don't make the RAID sets too large, and keep a hot spare or two in case of a disk failure. Maybe make two five-disk RAID-5 sets per 3310 and keep two disks per system in reserve. That would leave you with just over one TB of usable space, or 33% overhead.

Read Brett Cooper's answer to this question.

Dig Deeper on Data storage management