7200rpm + 5400rpm in RAID 1

You're better off just setting up the green as a backup and not as a component of the RAID1 setup. Putting the 5400 and 7200 in RAID will just hold back the performance of the 7200.

Edit:

I missed the part that you said you already had a backup. I dont recommend using a 5400 in RAID with a 7200 either way, it will just handicap the faster drive.
Edited by MightEMatt - 8/3/12 at 10:49pm

He said he doesn't have any backup.


The answer is no, don't build an array with disks of mismatched speeds, you would be shooting yourself in the foot. Instead, put both disks in the rig as stand alone, non member disks. Add a bat file to your scheduler to backup the content of the faster disk to the slower one every night at whatever time you are usually asleep. Problem solved.

i agree, make something like a robocopy bat file and set it to run in task scheduler plus you can just click on the .bat file any time you want to back up the 7200 HDD on to the 5400 HDD.. further info here http://www.overclock.net/t/1125994/seans-data-backup-guide

Unfortunately, using mismatched drives means both performance and longevity issues. Any time you have to HDDs that are not completely identical (they could both be 7200rpm, but different models from the same manufacturer for example) you will have performance degradation. Raid 1 (like all RAID levels with parity) requires that all data be written to both drives, and your hardware/software will have verify blocks of data (be reading it back from each drive and comparing the two) to ensure there have been no errors. If one drive has written a block and another one is not finished when the hardware/software goes to check, then it will obviously fail (this is a gross oversimplification, but you get the idea). When that happens you have to wait not only for the bock to be written, but then to be re-read, and checked against the other identical block for consistency. In practice this means that with mismatched drives you will never see the full performance of the slowest drive. That instead becomes your theoretical max write speed. Read speed is a little more tricky and will vary depending on your setup, but you might get the full read speed of your fastest drive.

I know sysadmins that don't even like to use identical drives from different batches for this reason. I recently bought 3 Seagate 3Tb drives for my HTPC and was surprised to see that 2 were made in China and one was made in Taiwan. Sure enough when doing my initial testing (I would never put data I want to keep on drives that had not been torture tested first) I saw about a 2-2.5% read/write penalty for using unmatched drives (with the same freakin model number). Luckily for me a client needed a 3Tb drive a few weeks later and I was able to quietly swap out for a 3rd unit matching the other two. Am I insane to worry about 2-2.5% difference? Probably. I'm I right to assume all the extra RAID block re-checking could result in earlier failure? Maybe, maybe not, but I'm not taking that chance.

To be fair though, with HDDs (as with every other hardware component) you're always playing the lottery. You can try to stack the deck in your favor, but there is no sure-fire way to know your hardware will last.

I have to do stress tests and risk assessments for network and server setups as part of my job, so I guess it's just a case of taking my work home with me. I would at the very least run a "long" SMART test. This is built into most Linux distros' disk utilities, but I don't have a clue how you would do it in Windows. I've always done it via the RAID card utility if I have to do that for a Windows server so my guess is you would need to get some type of 3rd party software. I would bet a quick Google search would yield several free options there. The "long" test takes a few hours but it actually scans the entire surface of each platter. The normal "short" SMART scan only shows bad sectors that have already been found in the course of normal operation. I guess you could format (full not quick) the drives and then see if any bad sectors were reported. Might work just as well.

just want to chime in that while everything posted above about identical drives is true, there is indeed a time to use mismatched drives. - They are less likely to physically break at the same time. It'll be slower, it may have more data errors if you're on a low-end controller, but you wont get two drives with a head crash in the same day.

For this reason and this reason ONLY, some users prefer to mismatch drives (and a very high quality controller) for times when eliminating down-time and data loss are more important than speed or drive longevity. Even then, it would still be similar drives (same rpm, interface, capacity, and disk number).

That's a brilliant insight. Plus 1! I can see this sort of thinking being appropriate for a RAID 1 (or probably more likely a RAID 5) array in a system that can only hold a small number of drives. RAID 6 solves the problem of multiple failures nicely, and RAID 10 would also get you most of the way there. RAID 10 also has the advantage of insanely fast read/write (4x and 2x). I'm getting off topic, but that's some serious outside-the-box thinking. I would suggest that this will cause more service disruptions (due to more frequent drive failures) in a production/everyday use array, but at least you'd be talking about having more frequent rebuilds (IE slow speed) and not full array failures (complete lack of access + restoring from the cloud or * shudder* tape).