[WBEL-users] WBEL kernel slower than RH9?

[Daniel T. Gynn]

--=-li29N7s5bsg8N+4m9rEs
Content-Type: text/plain
Content-Transfer-Encoding: quoted-printable

On Mon, 2004-03-08 at 15:20, Milan Kerslager wrote:
> On Mon, Mar 08, 2004 at 11:01:06AM -0500, Daniel T. Gynn wrote:
> >=20
> > When I was trying to narrow down a cause of a hardware problem I notice=
d
> > that the WBEL kernel seems slower than a RH9 kernel. =20
> >=20
> > I ran the following test to see for sure:
> > dd if=3D/dev/zero of=3D2gb bs=3D1024k count=3D2048
> >=20
> > date;copy 2gb 2gb.copy1;date
> > date;copy 2gb 2gb.copy2;date
> > date;copy 2gb 2gb.copy3;date
>=20
> time cp file1 file2
>=20
> > With kernel 2.4.21-9.0.1ELsmp, the difference between the dates was
> > 2:18, 2:23, 2:26.  With kernel 2.4.20-19.9smp, the difference between
> > the dates was 1:51, 1:49, 1:48.
>=20
> Make sure you have the same HDD settings:
>=20
> hdparm /dev/hda

I'm testing this on one server.  It has two SCSI drives setup as one
RAID 1 drive.  hdparm returns:

/dev/sda:
 readonly     =3D  0 (off)
 geometry     =3D 2212/255/63, sectors =3D 35544576, start =3D 0


>=20
> > This is on a Dual 2.8Ghz Xeon with 2GB RAM.  Nothing major is running
>=20
> Make the test in sigle user mode right after boot. Use Bonnie++ or
> something similar:
>=20
> http://www.coker.com.au/bonnie++/

Here are the results.  I set the hostname to the type of kernel.

Version  1.03       ------Sequential Output------ --Sequential Input- --Ran=
dom-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --See=
ks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec=
 %CP
wbel             4G 33043  98 42978  27 13565   5 27949  66 29353   6 419.4=
   1
                    ------Sequential Create------ --------Random Create----=
----
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Dele=
te--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec=
 %CP
                 16  2428  98 +++++ +++ +++++ +++  2553  99 +++++ +++  5549=
  99

Version  1.03       ------Sequential Output------ --Sequential Input- --Ran=
dom-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --See=
ks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec=
 %CP
rh9              4G 32446  97 42319  30 18097   7 29672  69 68666  15 534.1=
   2
                    ------Sequential Create------ --------Random Create----=
----
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Dele=
te--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec=
 %CP
                 16  2150  78 +++++ +++ +++++ +++  2886  99 +++++ +++  7036=
 100



It seems like the RH9 kernel uses a little more CPU and gets faster
times, whereas the WBEL kernel uses less CPU, but has slower times.


>=20
> > on the server yet to affect the times.  Does anyone have any ideas why
> > this is?
>=20
> There could be an issue with your HW. You did not told what type of HDD,
> IDE/SCSI you have.

I've also tested this on another server with the same results of the RH9
kernel being faster than the WBEL kernel.



--=20
-----------------------
Daniel T. Gynn
RHCE #806200978201621
Essential Systems, Inc.
412-931-5403 ext. 1
fax: 412-931-5425
dan.gynn@essensys.com
GnuPG Key http://www.essensys.com/~dan/gpgring.asc
Fingerprint: 0979 73B8 847A 349E 7363  66F4 6A79 DD72 495D CD60

--=-li29N7s5bsg8N+4m9rEs
Content-Type: application/pgp-signature; name=signature.asc
Content-Description: This is a digitally signed message part

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQBATkekanndckldzWARAqcwAKDBo/5A7U32aRqyKfWk71dd7u54kgCfQQDe
v93Nrh04PJd+e0kpcXnNp1g=
=pykj
-----END PGP SIGNATURE-----

--=-li29N7s5bsg8N+4m9rEs--