Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 24 Apr 2013 15:26:50 -0700
From:      "Marc G. Fournier" <scrappy@hub.org>
To:        Rainer Duffner <rainer@ultra-secure.de>
Cc:        "freebsd-fs@freebsd.org" <freebsd-fs@freebsd.org>
Subject:   Re: NFS Performance issue against NetApp
Message-ID:  <8A0622F0-D9FD-493B-8BDF-82D1359463F3@hub.org>
In-Reply-To: <F9A66A1B-4C3A-4E7A-AF8C-88B3C93D33A1@ultra-secure.de>
References:  <4F9A2AAE-938F-4FF5-A30C-72689D7F1F39@hub.org> <20130424093505.0f3e4415@suse3> <C643EE8D-5934-48EA-9305-EE6A07C5C897@hub.org> <C40EB001-4790-4BB9-A580-6D57D2CBC340@ultra-secure.de> <74CA0A5F-63EF-4B1B-B237-7A8DB76333A6@hub.org> <F9A66A1B-4C3A-4E7A-AF8C-88B3C93D33A1@ultra-secure.de>

next in thread | previous in thread | raw e-mail | index | archive | help

On 2013-04-24, at 14:18 , Rainer Duffner <rainer@ultra-secure.de> wrote:

>=20
> Am 24.04.2013 um 23:02 schrieb "Marc G. Fournier" <scrappy@hub.org>:
>=20
>>=20
>> On 2013-04-24, at 12:48 , Rainer Duffner <rainer@ultra-secure.de> =
wrote:
>>=20
>>>=20
>>> In the NFS mount options.
>>> I've re-done my tests and it still seems to be a problem.=20
>>=20
>> 'k, just to clarify =85 your tests are with the Broadcom chipset(s)?  =
 I have an Intel card on order, so will run tests against that as soon =
as it gets here to see  how it changes =85
>>=20
>=20
>=20
> No, with the intel card then.
> Even then it was slow.
>=20
>=20
>=20
>> Just curious, but what are you running for tests?   I noticed it =
visually, but other then  using something like iozone21, not sure what =
is useful for coming up with a benchmark =85
>=20
>=20
> I've just ran something like
> dd if=3D/dev/zero of=3Dbla bs=3D1000000 count=3D1000
>=20
> on the nfsmount on the client.
>=20

Here is what I'm seeing with the Broadcom card:

IOZONE performance measurements:
        108561574 bytes/second for writing the file
        95311340 bytes/second for reading the file


That is using 'ozone 1024' for a 1G file =85 and no mount / kernel =
settings =85

The problem isn't with the single file read / write though =85 the =
problem is when starting up an app server like jboss =85 my "reference =
server" is an old 32bit, 2 CPU + HT server with local drives =85 start =
up takes, from jboss start to deploy:=20

				JBoss AS 7.1.1.Final "Brontes" started =
in 87916ms

now, I do acknowledge that this is with local drives, so will always be =
faster the NFS drives =85 but =85=20

HP Proliant Gen8, 64bit, 16G of RAM, Oodles of CPU:
	FreeBSD/SMP: Multiprocessor System Detected: 24 CPUs
	FreeBSD/SMP: 2 package(s) x 6 core(s) x 2 SMT threads
Brand new HP Procurve 2910al-24G =85 brand new 3xxx series NetApp =
running 8.1.2 =85 and everything pretty much idle (non production yet), =
and:

				JBoss AS 7.1.1.Final "Brontes" started =
in 310249ms


So, over 4x slower?

Can you send me what your mount options are in /etc/fstab?  I'd like to =
try that out and compare against the two above =85 and I do have the =
Intel card on the way, so see if that makes a difference =85 =20







Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?8A0622F0-D9FD-493B-8BDF-82D1359463F3>