Making world but no kernel

[Chuck Swiger]

On Jul 26, 2011, at 8:03 AM, Jerome Herman wrote:
> On 26/07/2011 16:58, Chuck Swiger wrote:
>> On Jul 26, 2011, at 7:25 AM, Jerome Herman wrote:
>>> Actually it is Raid 10 of a sort. Three first halves of the three disk concatenated and mirrored on the three second half of the same drives.
>> There's a significant problem right there.  Not only will that configuration badly degrade the performance of the RAID volume, it also compromises the goal of redundancy which RAID-1 is supposed to provide.
>> 
>> Regards,
> 
> Disk are interweaved, so the performances are quite good (about 160% of a single drive)

A six-disk RAID-10 setup ought to provide nearly 600% read performance improvement and 300% of the write performance of a single drive-- real numbers tend to be perhaps 550%/250% or so.

> and the redundancy is here. Any single drive can fail, and the other two will be there to provide data.  Basically the first plesk is a-b-c, and the second is b-c-a, so everything should be fine.

Yes, if you do that you can survive a single disk failure, but the degraded performance is going to suck, and you have no chance of surviving a second disk outage.

A six-disk RAID-10 volume can survive up to 3 disks failing (although it has a 20% chance of losing the RAID with a two-disk failure and a 50% chance of losing data if a third disk goes without anyone fixing things).

Regards,
-- 
-Chuck