Re: AIX and RAW-device/filesystem

From: Hari Seldon <bill_at_trout.finance.state.mn.us>
Date: 26 Aug 1994 19:14:20 GMT
Message-ID: <33leud$drc_at_nic.dot.state.mn.us>


In <33d41q$hc9_at_watt.oedison.com>, rhodesr_at_watt.oedison.com (Richard L. Rhodes) writes:
>I've recently been informed I'm going to be part of our
>oracle spport (I'm currently AIX support). I just read
>the summary about this topic (I'm a latecommer to this
>news group). The Oracle installation book for AIX
>makes the two following statements:

 ..
>The manual appears to be an older version so is probably
>not accurate, BUT, I've been puzzling over this
>contradiction for some time.

the last time i communicated with the oracle porting team over the pond they indicated about a 2-5% gain. the oracle team, indead the individual who wrote the article about raw device drivers had never tried aix. the aix fs is really effeciant, and pretty fast. the use of std os files (imho) far outweighs the minimal performance gain on aix.

try this on your os.
go to the production machine.
while in use pull the plug
reboot
what does oracle say?
on aix with the journaled fs the worst i have ever seen it in the past 4 years is 'the database seems to have been stopped without shutting it down' (or some such) and thats it!. how do i know? we run aix hacmp and we tested the configuration (mostly in development, once by accendent in production) by unplugging things to see if the backup mechanism would work as planned. never an oracle crash, barely an fsck on reboot. so do you really want to putz with raw devices???

bill pociengel
dba/sa aix 3.2.4 hacmp dual networks (ie ethernet/tokenring)

so what is the wow wow wibble wobble wozzey woddle woo? 'so what's a flame but a polite message reduced to it's essence?' lilb pleceoing 'I'm a little confused right now' Received on Fri Aug 26 1994 - 21:14:20 CEST

Original text of this message