RAID 5 Array with different size drives.

[FireDemon]

I am about 99% sure that I know the answer to this, but I am just triple checking.

I plan on setting up a 1TB RAID 5 array with 3 500GB drives. One of the 500 gigers I had went kaput during diagnostics so I sent it back for an RMA. Seagate shipped me a brand new drive except when it arrived, they goofed and sent me a 750GB LOL! This drive is model ST3750640AS. The other two are models ST3500630AS which are basically identical except for the drive size.

Other than losing the extra 250GB to match the 500GB drives, will this affect performance any such as possibly different access times (Haven't checked the ms rating on the 750 yet) or different number of platters (That is if they added any in the 750 model instead of just increasing the platter size, haven't investigated this either).

 

[Cyrilix]

I believe in RAID 5, the speed defaults to the lowest of the two as it has to acquire data from each drive. Things probably aren't that simple, but here's a diagram explaining what I mean (T1, T2, T3, T4 represent different speed tests):

Drive A
T1 - [][][][][]
T2 - [][][]
T3 - [][][][]
T4 - [][][][]

Drive B
T1 - [][][]
T2 - [][][]
T3 - [][][][][]
T4 - [][][]

Overall Speed
T1 - [][][]
T2 - [][][]
T3 - [][][][]
T4 - [][][]

Of course, you still have to add in the fact that performance will increase relative to the "Overall Speed" diagram I have drawn, because of factors such as striping (reading from multiple drives). Speed will also decrease due to increased latency and parity calculation.

 

[Farva]

Raid will only use the smallest drive in array. So if you have a 40GB, 200GB, and 500GB, the array will only be able to use 40GB of each drive.

 

[volve]

From my numerous RAID setups, the difference in speed of the drives won't have any negative affects.

As you noted, the only apparent issue is the loss of 250GB, but that's not really a big deal is it?

 

[Farva]

From my numerous RAID setups, the difference in speed of the drives won't have any negative affects.

As you noted, the only apparent issue is the loss of 250GB, but that's not really a big deal is it?

Well, it is not recommended to run dissimilar drives in a Raid array anyway. Can you do it? Sure, but I surely wouldn't recommend it.

 

[FireDemon]

@ Protias,

Yes I am aware that it adjusts the array for the smallest drive, hence my saying "Other than losing the extra 250GB".

Also, what is the problem (Being seriously inquisitive here) about using different drives in this case? They all have the exact same interfaces, operating speeds, and latencies they are essentially the same drives except the 750 has more space. Also I suspect it has an extra platter because it weighs a tad more than my 500GB drives.

 

[volve]

@FireDemon:

It's simply a matter of latencies and timings. If the RAID controller issues a request and has to wait on 1 single drive longer than the others, it'll delay all of the subsequent queue requests, etc.

High-end storage systems run a series of tests on drives during initialization to establish a set of baselines when creating volumes so the timings can all be as pre-calculated and reliable as possible.

I imagine most of the desktop systems don't bother and are more tolerant of sporadic delays. As you said though, these drives are virtually identical even down to the controllers, so you'll likely notice little to no impacts at all.

 

[Farva]

@ Protias,

Yes I am aware that it adjusts the array for the smallest drive, hence my saying "Other than losing the extra 250GB".

Also, what is the problem (Being seriously inquisitive here) about using different drives in this case? They all have the exact same interfaces, operating speeds, and latencies they are essentially the same drives except the 750 has more space. Also I suspect it has an extra platter because it weighs a tad more than my 500GB drives.

You could run into problems with having different drives, but I do not remember enough about Raid to know if this is an issue or not.

All Seagate drives are 4 platters.