Re: RAID5 vs Mirrored drives with NT 4.0 server

But was that really pushing the I/O? What you state is contrary to eveything I have ever seen and read in all the Oracle books out there. And my own experience of RAID5 v non-RAID5 systems - in hardware as well as software.

I was at IBM last week, and their *heavy I/O* testing had RAID5 many, many times slower on write performance.

The best adivce I would give is TEST all of this in a REALISTIC environment for your own system and draw your OWN conclusions. Most *decent* RAID systems can be configured to RAID1, 0, 0+1 and 5, so it's worth spending some time playing around and getting some decent figures before you fix yourself with one way of doing things - because you might end up with a mix (say RAID5 for data and indexes if you don't need fast writes, and RAID1 - maybe with RAID0 striping - for the rest of your db. And maybe raw devices too for redo...)

Just my take.

Steve Phelan.

Jerry Gitomer wrote in message <351FE795.636E_at_ictgroup.com>...
>Hi,
>
>This subject came up at the DBA session of the most recent Delaware
>Valley Oracle User Group meeting. It was reported that using software
>RAID 5 slowed down writes by about 30%. It was also reported that when
>using hardware RAID 5 no loss was suffered.
>
>Regards
>
>Jerry
>
>
>bllag_at_aol.com wrote:
>>
>> I am seeking information regarding RAID 5 vs Mirrored drives on a server
>> running NT 4.0 and Oracle 7.3. I understand the economics, but I have
>> customers who continue to opt for cheaper rather than mirror. My
 company's
>> particular application is write intensive.
>>
>> Thank you for any help.
>>
>> -----== Posted via Deja News, The Leader in Internet Discussion ==-----
>> http://www.dejanews.com/ Now offering spam-free web-based newsreading
>
>--
>Jerry Gitomer ICT Group
>jgitomer_at_ictgroup.com Langhorne PA
>jgitomer_at_p3.net Opinions are mine not those of ICT Group
Received on Tue Mar 31 1998 - 00:00:00 CST