I have a RAID-5 Storage with 12 x 500GB drives. I have 1 faulty drive that needs to be replaced. The problem is I am unable to find the exact same part number drive. I would like to know if I can change this with another similar 500GB drive but with a different part number from the same manufacturer. I guess the new drive has a larger cache memory on the drive. Do I have to always use the same exact part number when building a RAID or can I use mixed drives ?
You can use any drive as long as the capacity is either the same, or larger (but not smaller). Note that two 160GB disks of different brands/series/types may have different capacities; one may be a few hundred megabytes larger than the other; this can create problems.
Also, if one drive is slower than the other, the slower drive may also affect performance of the other drives in some cases where the RAID is not properly setup. In properly configured arrays only the I/O handled by the slower disk would feel a slowdown, while the other (random) I/O would go at the usual speed on the other disks.
You should be aware that having such a large RAID-5 set is generally a poor idea. There are two problems that occur when you use a very large RAID-5 set:
1) When the set is degraded, reading a disk block that happens to lie on the drive that failed requires reading blocks from EVERY OTHER DRIVE in order to reconstruct the data. This can cause performance issues. And rebuilding such a large set will also take a very long time.
2) In order to recover the set, you have to be able to successfully read every block from the remaining 11 disks. Since hard drive reliability is typically quoted as one nonrecoverable read failure in every 10^14 bits read, and since a 1TB drive contains about 10^13 bits, 10 or more disks in a set is getting towards even odds that you won't be able to recover your data. When you combine this with the fact that with 12 drives you're 12 times as likely to experience a disk failure, you're asking for trouble. (Your drives are only half as large, so your odds of loosing data may be perhaps something like 1 in 4 instead of 1 in 2, but those are still pretty poor odds.)