Choosing between HDD's

Your links do not work. If the links are too long, the forum automatically shortens (truncate) them making them useless.

Do not paste links in the Quick Reply box. Click the Post Reply icon and from there look for and click the link icon which looks like a chain link.

Paste the link inside the two bracketed "URLs".

---------------------------------

To answer your question, yes you technically can, but it is not recommended. Different drives have different timings and specs. In the best case scenario you will take a performance hit because one drive has to wait for the other. In the worst case scenario the different timings can cause a skip in data writing which will render the data useless.
 

UncleDave

Distinguished
Jun 4, 2007
223
0
18,680
Seagate Barracuda 7200.10 ST3320620AS (Perpendicular Recording Technology) 320GB 7200 RPM 16MB Cache SATA 3.0Gb/s Hard Drive - OEM

Perpendicular Recording Technology

Manufacturer Warranty
Parts 5 years limited
Labor 5 years limited


As for mixing drives..... I can only re-iterate jaguarskx's post

To answer your question, yes you technically can, but it is not recommended. Different drives have different timings and specs. In the best case scenario you will take a performance hit because one drive has to wait for the other. In the worst case scenario the different timings can cause a skip in data writing which will render the data useless.
 

SomeJoe7777

Distinguished
Apr 14, 2006
1,081
0
19,280


I hardly think so. No error will ever arise from using mismatched drives unless there is a hardware problem.

Using mismatched drives will work fine with no errors. However, there is a performance degradation.

For example, say I build 3 different 2-drive RAID-0s, one with 2x drive A, one with 2x drive B, and a third with 1 each of A and B.

Let's say drive A has better and more intelligent cache handling, and therefore outperforms drive B with several requests for small files. But drive B has a higher sequential transfer rate (STR) and outperforms drive A when reading large files.

My first RAID-0 with 2x drive A performs much better than the second RAID with 2x drive B when reading several small files, thanks to the better caching in the A drives.

My second RAID-0 with 2x drive B outperforms RAID-0 with 2x drive A when doing Photoshop scratch work, thanks to the higher STR of the B drives.

The third RAID-0, with 1 each of drive A and B underperforms both of the other RAID-0s. In the several small files case, the RAID controller is always waiting on drive B to complete requests because it's caching isn't as good. In the STR applications, the RAID controller is always waiting on drive A to complete requests because of the slower STR. In short, the 1 drive of each RAID is always hampered by the worst-case performance of any single connected drive.

This is why it's generally recommended to use identical drives. Since the purpose of RAID-0 is to increase performance, and using different drives hampers performance, the use of different drives in RAID-0 is counterproductive.

For other RAID levels, like RAID-1, it's less counterproductive, but the performance still suffers.
 

UncleDave

Distinguished
Jun 4, 2007
223
0
18,680


Can we agree that it is recommended that all drives in a RAID array SHOULD be the same but can be different?

Next time I quote I'll make sure I quote properly not cut and paste - If data ever got (gets?) out of sync between two drives it's up to the controller card (or software) to correct it.
 

SomeJoe7777

Distinguished
Apr 14, 2006
1,081
0
19,280


Yes, of course. That is for sure.



I am confused on this point. What does it mean that data "gets out of sync" between the two drives? Why would that happen? And why would different drives make this more likely?

The RAID controller is in absolute control of all data writes and reads. If a drive does not do what the controller has told it to do, then the drive has failed for all intents and purposes. You now have a degraded RAID (or useless RAID-0), and must replace the drive.
 

UncleDave

Distinguished
Jun 4, 2007
223
0
18,680


I think we are saying the same thing.

The RAID controller is not in absolute control of all data writes and reads - the controller instructs the drives to store or retrieve information. The drive will let the RAID controller know the results of the instruction. The read or write request can fail for any number of reasons from the kick I just gave the server to the cable expanding and making dodgy contact to any number of reasons. If the RAID controllers request succeeds on the first drive and fails against the second drive, the RAID controller has a number of actions to take, the last of which will result in the drive being disabled.

I have never contended that it is more or less likely - I quoted somebody else <--- my mistake. IMHO Data can get out of sync for the hell of it :pt1cable:

I certainly do not contend that the RAID controller will detect and then just leave data of of sync.
 

SomeJoe7777

Distinguished
Apr 14, 2006
1,081
0
19,280


To me, data would only be in a state that the RAID controller does not desire and/or did not intend if there is an error condition happening somewhere downstream of the RAID controller. Yes, all of those reasons you mentioned can cause a drive to have trouble writing or reading data. However, the drive's own internal algorithms will re-try the operation, attempt ECC, etc. before reporting to the RAID controller that the command did not complete. By the time a drive reports an error to the RAID controller, there is a serious problem.

During that entire period of waiting on the drive to re-try, the RAID controller is patiently sitting there, waiting for a response. The RAID controller will never give up and leave the data in an unknown state (barring a software/firmware bug in the controller). Once the drive actually throws an error condition back to the controller, the controller can take several actions, as you've stated, including it's own attempts at re-trys, or if the array is redundant (RAID-1 or 5), attempt to immediately reconstruct the data from other drives, and/or reallocate the bad block. It can and will also warn the user through the software driver and user interface.

Anyway, my point here is that to me, if any of this type of error correction is coming into play during RAID operation, the array is not functioning properly from a hardware perspective. There should never be a situation where active error correction steps should be considered part of normal operation. I don't believe that data "gets out of sync" for the hell of it. There is always a reason. That reason is usually defective hardware, and that defect needs to be fixed.
 

vaker5

Distinguished
Oct 28, 2007
136
0
18,680
Whats the point of buying a bunch of different drives anyways why not get all the same... is there a point, if not then why do it?
 

UncleDave

Distinguished
Jun 4, 2007
223
0
18,680


You having a laugh?
 

prolfe

Distinguished
Jan 9, 2005
252
0
18,780
This has been a very informative post for me; thanks! I recently experimented with RAID0 and couldn't figure out why two drives were slower than one (they were VERY dissimilar, different brands, capacities, ETC).