Raid definitions

[Lord of Gore]

Stephen Carville wrote:
> Lord of Gore wrote:
>> McDougall, Marshall (FSH) wrote:
>>> I am in the process of allocating drive space in several new servers.
>>> Each of these servers are to house DB's of some sort.  Each has 6 X 146
>>> gig drives and each will run RHES4.  I was thinking of 2 arrays:
>>>
>>> 1 - mirrored pair of 146's for OS
>>> 1 - Raid 5 array with 4 x 146 for db's.  That would give me a disk of
>>> about 440 GB.
>>>
>>> Currently, the biggest DB is about 30 GB.   I figure there are just too
>>> many ways that I can slice and dice it.  Any advice appreciated.
>>> Thanks.
>>>
>>> Regards, Marshall
>>>   
>> I'd go with level 5 for the extra space. Fault tolerance is the same: 
>> just 1 broken disk per "session" :) .
>>
>
> A four disk RAID 10 set can lose up to two drives without failing as 
> long as both are not in the same mirror set.  
Seems to me that I have to copy paste definitions from e-books. Of 
course in RAID 10 bla bla bla and the guy seems to know a little about 
these things, there's no need to clog this thread with useless info. I 
said I'd go for lvl 5 for the space. Anyway I think that bottlenecks are 
more likely to occur at hdd space level instead of hdd access time. But 
these are questions that should be answered by mr. McDougal. So, mr 
McDougal what kind of work will do these db's of yours? Give us a little 
more infos on the subject. We've got ourselves a very meticulous reader. 
While at it see how much it would cost to expand the array. It's a 
meticulous question.
> RAID 10 has much better read and write performance than RAID 5. Should 
> a drive fail, rebuilds to the replacement drive are faster making the 
> window for a catastrophic failure less.
>

Well indeed it seems that although unnecessary we should dig a lot more 
and give mr Marshall a headache and suggest him that he should use 
drives from different manufacturers to lower the possibility to have 2 
drives fail in the same time. But I have a strong feeling that he has 
all the disks from the same manufacturer. So given the fact that they 
are manufactured in approximately the same time they are subjected to 
the same stress and therefore fail in approximately the same time.
In this case he should forget about servers and in fact about the whole 
IT thing and go to a monastery. Or he should try enterprise solutions 
like RAID 6 or even better 0+1+5 or 0+1+6 if hardware permits it and 
hope that the giggly girl from PR that he likes so much (no offense mr 
Marshall I'm just using her to make a point :) ) and pays him every once 
in a while a visit while bringing him a cup o coffee doesn't spill it on 
the servers.
Or, and I think this is the best solution and the *complete* answer, we 
could let Marshall know what *we* as singular individuals would opt for 
and let him choose and forget about posing in the "know it all" Batman 
arret.

Cheers