Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: 9i multi cache buffer

Re: 9i multi cache buffer

From: Nuno Souto <nsouto_at_optushome.com.au.nospam>
Date: Wed, 13 Nov 2002 21:42:27 +1100
Message-ID: <3dd22f35$0$4367$afc38c87@news.optusnet.com.au>


Wed, 13 Nov 2002 12:24:37 +1100, Howard J. Rogers said (and I quote):

> My advice on such matters is (1) consult your vendor and (2) consult Steve
> Adams, who says that HP-UX has an 8K block size. He goes on to say the
> following (which makes about as much sense to me as the output from top):

In most Unix "flavours" you can set a preferred block size in cooked file systems. The default varies indeed between 4K and 8K, but it can be any multiple of 512 between 1024 and 32K, IIRC.

Of course it makes sense to not stray too far from other internal constants, such as page size and such. Which again vary between 4K and 8K (any wonder why the default file system block size is one of those two values? Hint: check the page size... It's all to do with memory management and sharing paging algorithms).

"sar -d" is Unix lingo for "monitor disk I/O". blks/s is file system blocks per second, r+w/s is raw read+write requests to the disk per second. Usually you can work out from these a ratio of physical to logical I/Os (within the file system) and based on the overall I/O byte rate figure out the block size.

There are also a number of commands that will dump the file system block size, but they vary wildly between Unix versions.

Of course, none of this is too much of a prob for you privileged NT people... ;)
Then again, watch out for that sector size during format!

-- 
Cheers
Nuno Souto
nsouto_at_optusnet.com.au.nospam
Received on Wed Nov 13 2002 - 04:42:27 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US