[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ale] Gigabyte MoBo RAID and RH-7.2



I just checked some Google sources to be sure, and they're all saying 
that RAID 0 does give a performance improvement.  This makes a certain 
amount of sense to me because you're reading and writing "across" drives 
and as long as drive buses are faster than drives, you're going to be 
better off.  I agree that with drives getting plenty fast, you probably 
can't expect an eight-drive RAID 0 array to be a big improvement over, 
oh, a four-drive array - but if you did it anyway, you'd have one 
accident-prone mass-storage resource on your hands! :-)

Adaptec's info gives RAID 1 (mirroring) mad props on random writes.  

When I worked for the Dept. of Energy, I had the use of a 6xPPro/200 ALR 
server with a three-channel RAID controller that ran about $2000 (the 
controller).  Man, I miss that thing.  After I left, no one could get it 
to work, so they got rid of it.  Because you had more than one channel 
(the card had three Intel i960 CPUs on it), you could split your mirrors 
up onto not just separate drives but separate channels.  That helps you 
guard against controller failures and speeds you up somewhat.

- Jeff


Michael Smith wrote:

>My average IO speed is a good bit higher using RAID0 than no RAID at all.  
>I forgot what the command was to check the speed(someone posted this a 
>month or so ago) but when I compared it to others postings, my io speed was 
>a good bit faster.  I guess its comparing apples to oranges unless you have 
>the same Mobo and hard drives, though.
>
>Mike
>
>
>>On Wed, Mar 13, 2002 at 11:44:58PM -0500, Jeff Hubbs wrote:
>>
>>>Maybe that's why no one has busted buns to write the driver; why do it
>>> when kernel RAID is sitting there?
>>>
>>...especially when the kernel raid has been extensively tested ans is
>>known to be extemely reliable.  :)
>> 
>>You can't hot-swap drives with the onboard controllers, you can't add
>>and remove drives to the array (on the fly) with the onboard
>>controllers.. etc.
>>
>>>Yeah, that's true, but there may be times when you want "ludicrous 
>>>speed" from your drives and you're willing to get it at the expense of
>>> reliability.  Thankfully, many RAID controllers (dunno if kernel RAID
>>> does this yet) will let you make stripe sets out of mirrors or vice 
>>>versa such that you can get that speed and still be able to lose one
>>>or  more drives.
>>>
>>Oh, the kernel has let you do that for quite some time now.   RAID5 a
>>set of RAID5 arrays?  No problem.  :)
>>
>>But in all seriousness, single drive transfer rates are exceeding 40
>>MB/sec, so you're rarely hurting for sustanied transfer speeds.  RAID0
>>doesn't boost average access times either -- because you're striping,
>>both drives generally need to be doing the same thing at the same time
>>so you can't have 'em seeking individually -- and that's what really
>>boosts performance.  Average I/Os per second.  
>>
>> - Pizza
>>-- 
>>Solomon Peachy                                   
>>pizzaATfucktheusers.org I ain't broke, but I'm badly bent.             
>>             ICQ# 1318344 Patience comes to those who wait.
>>    ...It's not "Beanbag Love", it's a "Transanimate Relationship"...
>>
>
>




---
This message has been sent through the ALE general discussion list.
See http://www.ale.org/mailing-lists.shtml for more info. Problems should be 
sent to listmaster at ale dot org.