Re: RAID experience/recommendations

In article <slrn6b9dmu.qth.oracle_at_tchp2.tcamuk.stratus.com>, Neil Chandler <oracle_at_tchp2.tcamuk.stratus.com> writes
>In article <SEKaeAAwB0s0Mw6r_at_camk.demon.co.uk>, Andy Hardy wrote:
>
>Redo logs are written sequentially. If possible, it is better to keep them
>away from your other datafiles which are written randomly and thus keep
>the disk head away from where you would like it to be (i.e. the disk head must
>keep returning to the last point written in the Redo).

Ah... very interesting point. Thanks.
>
>Ideally, Redo mirrored by Oracle on 2 separate 'ordinary' drives is my
>prefered choice. I have also found that triplicating Redo logs across 3 drives
>is 20% slower that duplicating (but I don't know why yet).
>
>Raid 0+1 is the best (stripe+each disk mirrored) if you can afford it,
>but you have a 50% disk redundancy.
>Raid-5 gives a 33-20%-ish redundancy but is considerably slower - especially
>should you lose a drive and have to re-sync the raid set on-line.
>
>You pays your money, you takes your choice.
>
>regs
>
>Neil Chandler
>
>
>>In article <19980107021500.VAA17428_at_ladder02.news.aol.com>, BRAJB
>><brajb_at_aol.com> writes
>>>I don't remember if there is a white paper or not, but we are using HP 0+1 in
>>>our information warehouse and will be in a few weeks implementing it on our
>>>transaction system. The key concepts that have driven home for me are the
>>>following:
>>>
>>>1. Hardware stripe not software
>>>2. Redo logs on other that stripe set, ours are on the internal system disk.
>>>
>>>Hope this helps.
>>Why the second?
>>
>>Andy
>>--
>>Andy Hardy
>>Senior IT Systems Engineer
>>Cegelec AEG
>>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>Opinions are mine and do not necessarily reflect those of Cegelec AEG
>
 

-- 
Andy Hardy
Senior IT Systems Engineer
Cegelec AEG
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Opinions are mine and do not necessarily reflect those of Cegelec AEG