Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.server -> Re: Raw partitions vs. file system
Billy Verreynne <vslabs_at_onwe.co.za> wrote in article
<01bce435$c63b35a0$f3040059_at_billyv.vslabs.co.za>...
> S V <sv1_at_mindspring.com> wrote in article
> <01bce3d4$43dd9be0$db5c0c26_at_sfinance3>...
>
> > I have got quite significant perfomance boost when I went FROM
> > filesystem TO raw logical volumes on AIX 4.2 platform.
> > And, in fact, Oracle perfomance tuning guide/tips for AIX platform
> > does claim that Oracle performs significantly and consistently better
> > when using raw devices than when using files.
>
> This is interesting. We went thru a similar excercise more than a year
ago
> and finally decide on using file systems (easier to administrate and move
> around for i/o load balancing) on a SMP box. At that time the docs for
> Oracle for that platform suggests not using raw devices due to the
> complexity of raw vs cooked files.
this "complexity" evaporates as soon as you learn how to use "dd". Having said what I said about using raw devices, I may actually back off a little bit and mention that in _some cases cooked files perform better than raw logical volumes. This happens when system takes advantage of read-ahead on the file system. This can give some imporvement on full table scans, however at a price of more memory consumption for file buffering which can be better used by Oracle for block buffering on raw devices.
> My initial reaction was to use raw devices, but after reading the docs
and
> talking to local Oracle, the company decided to use cooked files instead.
I
> was under the impression that there should be some noticeable increase in
> performance when using raw devices.
>
> > Furthemore I found them to be easier to administer than filesystems.
> > Just compare the time it takes to create 2G logical volume ( < 1 min)
> > with a time it takes to create a filesystem the same size.
>
> What about backups? To backup raw devices in Unix (when you don't have
any
> special backup software) you need to use dd - and I won't rely on dd
> backups for a database.
what's wrong with "dd" ???
I would rather rely on dd which is absolutely
reliable piece of soft furniture, tested over generations than on some
third-party tools full of bugs.
> Also with file systems it's easy to move Oracle
> datafiles around between mount points to spread the i/o load - this is
not
> possible with raw devices which makes fine-tuning of disk i/o performance
a
> bit difficult.
well, this is simply NOT TRUE.
All it takes
to move a datafile from /dev/rlv01 to /dev/rlv02
is
ALTER TABLESPACE AAA OFFLINE;
dd if=/dev/rlv01 of=/dev/rlv02 bs=4k skip=1 seek=1
ALTER TABLESPACE AAA RENAME DATAFILE '/dev/rlv01' TO '/dev/rlv02';
ALTER TABLESPACE AAA ONLINE;
It's no more difficult than to do the same operation on cooked file,
and it's faster too.
> But then it also depends on the hardware. On another
> platform (MPP) we're using raw devices. If a disk controller/channel is
> suffering, you simply pull and swap physical disks (even between
different
> nodes and even if the disk is part of a RAID set) to balance the i/o
load.
> Not bad! :-)
>
> > Besides,
> > it's quite easy to delete by mistake an Oracle file on a filesystem
> > (we have seen quite a few folks here begging for help after doing
> > just that). Not so for raw logical volumes.
>
> Sheez - this is dangerous. But then it means that too many people (and
the
> wrong people) have root and oracle access to that box.
to delete raw device it takes _at least root access to the box. I would never use root account for running a database. To delete a cooked file, it suffices to be only oracle, which is a step lower level of security.
Sincerely,
Serge
> Or that the file
> permissions of the datafiles are incorrect. Also a proper directory
> structure makes administration easier, especially if you're running
> multiple databases on the same box.
>
>
> regards,
> Billy
>
>
Received on Wed Oct 29 1997 - 00:00:00 CST
![]() |
![]() |