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: multiblock read count question

Re: multiblock read count question

From: Rob K <robkato_at_earthlink.net>
Date: Wed, 20 Jun 2001 19:18:23 GMT
Message-ID: <3G6Y6.22725$aV1.2257238@newsread1.prod.itd.earthlink.net>

I believe the filesystem I/O is buffered using Veritas. Is there an easy way to find out?
"Yong Huang" <yong321_at_yahoo.com> wrote in message news:b3cb12d6.0106201048.12d3a771_at_posting.google.com...
> This is what Steve Adams says:
>
> This is what the Ixora web tip on multiblock reads at
> http://www.ixora.com.au/tips/creation/multiblock.htm says on the matter
 ...
>
> "Unfortunately, if you use buffered I/O the operating system will service
> multiblock reads as a series of small I/O operations anyway and the only
> benefit
> from multiblock reads will be a small saving in CPU usage. In this case
 the
> multiblock read count should be approximately equal to the typical track
 length
> to take advantage of track buffering and no larger lest the optimizer be
> deceived."
>
> The key word there is "buffered". If your filesystem is able to do direct
 I/O,
> and is configured to do so, then the large multiblock read principle
 applies.
> But, if your filesystem I/O is buffered, then there is very little to
 gain.
>
Received on Wed Jun 20 2001 - 14:18:23 CDT

Original text of this message

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