4 Drives - What kind of raid array?

Main Target

[H]ard|Gawd
Joined
Jul 5, 2004
Messages
1,690
So I have 4 matching Samsung 750 gig drives and I was toying around with what kind of raid array I want to run it on.

I was thinking either:
Raid 0 System
Raid 1 Storage

Or

Raid 0+1

Just a note space isnt a huge concern here. I am curious however about how effective raid 01 is. Has anyone here used it? How does it compare to that of straight up raid 0?

Thanks for your help guys :)
 
Well, with RAID0, if you lose a drive then all your data's gone.
I'd setup a RAID5 myself. That way if you lose a drive, you don't lose everything, just replace the drive, and off you go.
 
Isnt raid 0+1 just a raid 5 with 1 more drive?

RAID 5
Striped set with distributed parity or interleave parity requiring 3 or more disks. Distributed parity requires all drives but one to be present to operate; drive failure requires replacement, but the array is not destroyed by a single drive failure. Upon drive failure, any subsequent reads can be calculated from the distributed parity such that the drive failure is masked from the end user. The array will have data loss in the event of a second drive failure and is vulnerable until the data that was on the failed drive is rebuilt onto a replacement drive. A single drive failure in the set will result in reduced performance of the entire set until the failed drive has been replaced and rebuilt.

RAID 10
RAID 1+0 (or 10) is a mirrored data set (RAID 1) which is then striped (RAID 0), hence the "1+0" name. A RAID 1+0 array requires a minimum of four drives – two mirrored drives to hold half of the striped data, plus another two mirrored for the other half of the data. In Linux, MD RAID 10 is a non-nested RAID type like RAID 1 that only requires a minimum of two drives and may give read performance on the level of RAID 0.

RAID 01
RAID 0+1 (or 01) is a striped data set (RAID 0) which is then mirrored (RAID 1). A RAID 0+1 array requires a minimum of four drives: two to hold the striped data, plus another two to mirror the first pair.
 
RAID 10 can lose 1 or more drives so long as neither is in the same mirror. You cannot lose more than one drive in a RAID01 set.

We're converting all of our stuff from RAID5 to RAID10. RAID5 made sense when drive capacities were small and larger drives were expensive.

RAID5 write times are worse than RAID10. The rebuild times for a degrated RAID5 array are much longer than a similar RAID10 array.

Performance of a degraded RAID10 over a RAID5 array are just as great, and that's both just degraded performance as well as while both are trying to rebuild.

RAID5 sucks unless you don't have the money and/or expansion bays.
 
Incorrect. Both can sustain multiple drive failures depending on which drives they are.

Yep, you're right, BUT if with RAID10 you can suffer a loss in both sets, RAID01, you can't.


4 drives RAID 01


-------------------------MIRRORED SET-----------------------

STRIPE1 STRIPE2

DRIVE1 DRIVE2
DRIVE3 DRIVE4

You can't lose a drive in both stripes

4 drives RAID 10:


-------------------------STRIIPED SET-----------------------

MIRROR1 MIRROR2

DRIVE1 DRIVE2
DRIVE3 DRIVE4


You can lose 1 drive in MIRROR1 and 1 drive in MIRROR2 and still
have your array functional.

RAID10 is also more robust that RAID01

http://www.pcguide.com/ref/hdd/perf/raid/levels/multXY-c.html
 
Last edited:
RAID 0+1 and RAID 1+0 both have guaranteed protection against one drive failure, and potentially up to 50% failure depending on which disk. So their resilience against failures is the same.

However, striping two mirrors would make it potentially more robust because only one mirror would have to be rebuilt when a drive failure occurs. Where otherwise, there would be one mirror with each a striped configuration - a rebuild would involve all disks.
 
RAID 0+1 and RAID 1+0 both have guaranteed protection against one drive failure, and potentially up to 50% failure depending on which disk. So their resilience against failures is the same.

The second sentence is incorrect.

Consider an N-drive RAID 10. Also, assume for the sake of argument that all drives are equally likely to fail (not a good assumption when one drive is being fully read to rebuild another, but leave that aside for now). After 1 drive fails, if the other drive in the mirror fails there will be data loss. But if any of the other N-2 drives fail, there is no data loss. So the probability of another drive failure causing data loss is 1/(N-1).

Now consider an N-drive RAID 0+1. After 1 drive fails, then any failure in the surviving N/2 stripe set will cause data loss. So the probability of another drive failure causing data loss (N/2)/(N-1).

Note that the second probability is larger than the first as long as N/2 > 1, or N > 2, which will always be the case for N-drive RAID 10 or 0+1.

As for the "bad" assumption earlier, while it affects the overall probability of data loss (making both higher than computed above), it affects both probabilities in a similar manner, so the conclusion still holds -- RAID 10 is more robust than RAID 0+1.
 
Last edited:
Yes you're right, but the difference is not all that exciting:

RAID1+0
In a RAID1+0 where two mirrors are striped, if one drive fails:

DRIVE A - stripe1
DRIVE B - stripe1 (mirror)
DRIVE C - stripe2
DRIVE D - stripe2 (mirror)

If disk A fails, disks C or D may fail but not disk B.
If disk B fails, disks C or D may fail but not disk A.
If disk C fails, disks A or B may fail but not disk D.
If disk D fails, disks A or B may fail but not disk C.

RAID 0+1
Now let's have a look at RAID0+1 (2 striped arrays are mirrored), if one disk fails:

DRIVE A - mirror1
DRIVE B - mirror1
DRIVE C - mirror2
DRIVE D - mirror2

If disk A fails, disk B may fail, but not disks C or D.
If disk B fails, disk A may fail, but not disks C or D.
If disk C fails, disk D may fail but not disks A or B.
If disk D fails, disk C may fail but not disks A or B.

In other words, you are guaranteed against one failure, and have a chance to cope with a second failure as well. With RAID 1+0, you have a slightly higher chance to still be accessible after 2 drive failures; but then you're gambling with your data. The benefit is also purely theoretical. Most RAIDs fail because of a RAID-specific issue, not because HDDs die. Especially windows fakeRAIDs are very sensitive and break easily; losing your data due to some RAID-engine issue is more likely than actually having a crashed HDD.

For home users, mirroring is a poor solution. As i understand, the topicstarter would not use a backup, but rely solely on the RAID; in fact using RAID as if it were an easy backup solution; except it's not.
 
In other words, you are guaranteed against one failure, and have a chance to cope with a second failure as well. With RAID 1+0, you have a slightly higher chance to still be accessible after 2 drive failures; but then you're gambling with your data. The benefit is also purely theoretical. Most RAIDs fail because of a RAID-specific issue, not because HDDs die.

With N=4, the chance of data loss after losing 1 drive is about double with RAID 0+1 as compared to RAID 10. I'd call double more than "slightly higher". With N=8, the same chance is quadruple. In general, it is N/2.

As far as most RAID failures not occurring because of an HDD failure, that contradicts my experience. Certainly it is possible to put together a RAID system that has incompatible hardware or is misconfigured, but anyone who takes the time to choose their hardware, set it up properly, and test it, can build a system with a very small chance of having the non-HDD RAID system fail.
 
If you controller supports it, RAID6. If not, then RAID5 + 1 spare.

RAID 6 with 4 drives is a possibility, if (as you say) his system supports it, and he cares more about robustness than speed.

But a 3 drive RAID 5 plus hot spare would be a poor choice. That would have the same capacity as a 4 drive RAID 10, but would be slower and more likely to lose data.
 
Back
Top