Re: [tlug] Linux CVS kernel tree in Japan

["A.Sajjad Zaidi" <sajjad@example.com>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Thu, May 30, 2002 at 04:22:55PM +0900, Jean-Christian Imbeault wrote:
> 
> I would but the reason I was asking for cvs was so that I would only 
> download whatever new files there were and not the whole thing all over 
> again.

Thats what patches are for too.
 
> Patch the kernel tree? Something else I have to learn ;) I was happy using 
> CVS for the 2.4-xfs tree since there no "patching" needed. Any new/modified 
> files were automatically download with a 'cvs -update'.

Its the same as applying the XFS patch mentioned in an earlier thread,
although its better to wait for the release unless you *have* to have it
now.

> Will try Reiserfs for my next server and do some benchmarking to see which 
> is best fo me.

That would be interesting, but remember that for any meaningful results,
you have to have a very controlled environment. That is, exactly the same
hardware, kernel version, block size, file size, file number etc.

You might save a lot of time and effort by looking at some existing
benchmarks and tools.
 
> Oh, and I made a mistake in saying that my IDE HD speed was 2.5 times 
> faster with a newer kernel. Actually it's only about 40% faster. The 2.5 
> times increase ocures if I use a new kernel *with* an XFS fs. What a 
> difference an XFS fs does in terms of speed! (I only used hdparm for 
> benchmarking, so take that with a big grain of salt).

I doubt changing the file system alone would result in a 2.5x change and
besides, hdparm tests the raw disk, not the file system. I suspect that
its the settings related to your chipset that are making the difference.

Run hdparm without any parameters on /dev/hda or whatever your HD is and
see what I/O support and DMA are set to.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)

iD8DBQE89eKPt1KjqyZ+DQ4RAkEWAJ426i1wIg7xBU6puQrnBYROh2oa2wCgq815
CnDCNFA+3Cx+5/YfgUAVxh8=
=FPD9
-----END PGP SIGNATURE-----