[UPDATE] Fresh zerocopy patch on kernel.org

[UPDATE] Fresh zerocopy patch on kernel.org

[David S. Miller]

At the usual place:

ftp://ftp.kernel.org/pub/linux/kernel/people/davem/zerocopy-2.4.1-1.diff.gz

(As usual, please allow some minutes for the mirrors to get it)

Changes since last installment:

1) Merge to 2.4.1 final. (me)
2) Accept TCP flags (ACK, URG, RST, etc.) for out of window packets
   if truncating the data to the window would make that packet valid.
   (Alexey)
3) Add SO_ACCEPTCONN, Unix standard wants it. (me)

Have fun testing...

Later,
David S. Miller
davem@redhat.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[Chris Wedgwood]

On Tue, Jan 30, 2001 at 01:33:34AM -0800, David S. Miller wrote:

    2) Accept TCP flags (ACK, URG, RST, etc.) for out of window packets
       if truncating the data to the window would make that packet valid.
       (Alexey)

    3) Add SO_ACCEPTCONN, Unix standard wants it. (me)

these have been feed back for 2.4.x Linus anyhow right?


  --cw
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[David S. Miller]

Chris Wedgwood writes:
 > On Tue, Jan 30, 2001 at 01:33:34AM -0800, David S. Miller wrote:
 > 
 >     2) Accept TCP flags (ACK, URG, RST, etc.) for out of window packets
 >        if truncating the data to the window would make that packet valid.
 >        (Alexey)
 > 
 >     3) Add SO_ACCEPTCONN, Unix standard wants it. (me)
 > 
 > these have been feed back for 2.4.x Linus anyhow right?

Yes, but I couldn't get them to him in time for 2.4.1

Later,
David S. Miller
davem@redhat.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[Tom Rini]

On Tue, Jan 30, 2001 at 01:33:34AM -0800, David S. Miller wrote:

> Have fun testing...

I've got a question.  I think one of the threads mentioned that this could
help NFS performance, in general.  I assume having the NFS traffic go out
the a card w/ the right HW could help as well, right?  Should the 3com
3c905b work, and how can i check it's "working" ?  Thanks!

-- 
Tom Rini (TR1265)
http://gate.crashing.org/~trini/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[Malcolm Beattie]

David S. Miller writes:
> 
> At the usual place:
> 
> ftp://ftp.kernel.org/pub/linux/kernel/people/davem/zerocopy-2.4.1-1.diff.gz

Hmm, disappointing results here; maybe I've missed something.

Setup is a Pentium II 350MHz (tusk) connected to a Pentium III
733MHz (heffalump) (both 512MB RAM) with SX fibre, each with a
3Com 3C985 NIC. Kernels compared are 2.4.1 and 2.4.1+zc
(the 2.4.1-1 diff above) using acenic driver with MTU set to 9000.
Sysctls set are
    # Raise socket buffer limits
    net.core.rmem_max = 262144
    net.core.wmem_max = 262144
    # Increase TCP write memory
    net.ipv4.tcp_wmem = 100000 100000 100000
on both sides.

Comparison tests done were
    gensink4: 10485760 (10MB) buffer size, 262144 (256K) socket buffer
    ftp: server does sendfile() from a 300MB file in page cache,
         client does read from socket/write to /dev/null in 4K chunks.

                       2.4.1                    2.4.1+zc
                     KByte/s tusk%CPU heff%CPU  KByte/s tusk%CPU heff%CPU
gensink4
  tusk->heffalump    94000   58-100   93        54000   98-102   11-45
  heffalump->tusk    72000   86-100   46-59     70000   71-93    53-71

                      2.4.1                     2.4.1+zc         
                      KByte/s                   KByte/s
ftp heffalump->tusk   86000                     62000


I was impressed with the raw 2.4.1 figures and hoped to be even more
impressed with the 2.4.1+zc numbers. Is there something I'm missing or
can change or do to help to improve matters or track down potential
problems?

--Malcolm

-- 
Malcolm Beattie <mbeattie@sable.ox.ac.uk>
Unix Systems Programmer
Oxford University Computing Services
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[David S. Miller]

Malcolm Beattie writes:
 > David S. Miller writes:
 > > 
 > > At the usual place:
 > > 
 > > ftp://ftp.kernel.org/pub/linux/kernel/people/davem/zerocopy-2.4.1-1.diff.gz
 > 
 > Hmm, disappointing results here; maybe I've missed something.

As discussed elsewhere there is a %10 to %15 performance hit for
normal write()'s done with the new code.

If you do your testing using sendfile() as the data source, you'll
results ought to be wildly different and more encouraging.

Later,
David S. Miller
davem@redhat.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[Malcolm Beattie]

David S. Miller writes:
> 
> Malcolm Beattie writes:
>  > David S. Miller writes:
>  > > 
>  > > At the usual place:
>  > > 
>  > > ftp://ftp.kernel.org/pub/linux/kernel/people/davem/zerocopy-2.4.1-1.diff.gz
>  > 
>  > Hmm, disappointing results here; maybe I've missed something.
> 
> As discussed elsewhere there is a %10 to %15 performance hit for
> normal write()'s done with the new code.
> 
> If you do your testing using sendfile() as the data source, you'll
> results ought to be wildly different and more encouraging.

I did say that the ftp test used sendfile() as the data source and
it dropped from 86 MB/s to 62 MB/s. Alexey has mailed me suggesting
the problem may be that netfilter is turned on. It is indeed turned
on in both the 2.4.1 config and the 2.4.1+zc config but maybe it has
a far higher detrimental effect in the zerocopy case. I'm currently
building new non-netfilter kernels and I'll go through the exercise
again. I'm confident I'll end up being impressed with the numbers
even if it takes some tweaking to get there :-)

--Malcolm

-- 
Malcolm Beattie <mbeattie@sable.ox.ac.uk>
Unix Systems Programmer
Oxford University Computing Services
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[David S. Miller]

Malcolm Beattie writes:
 > Alexey has mailed me suggesting the problem may be that netfilter
 > is turned on.

Oh yes, netfilter being enabled will cause some performance
degradation, that is for sure.

Later,
David S. Miller
davem@redhat.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[Pau Aliagas]

On Tue, 30 Jan 2001, David S. Miller wrote:

> Have fun testing...

Is it of any use testing it on a Xircom Tulip Cardbus card?

Pau


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: [UPDATE] Fresh zerocopy patch on kernel.org

[Chris Wedgwood]

On Tue, Jan 30, 2001 at 04:58:38PM +0100, Pau Aliagas wrote:

    Is it of any use testing it on a Xircom Tulip Cardbus card?

NFS might seem an improvement because of delayed fragmentation; other
than that -- no point



  --cw
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: [UPDATE] Fresh zerocopy patch on kernel.org

[Jonathan Earle]

> Malcolm Beattie writes:
>  > Alexey has mailed me suggesting the problem may be that netfilter
>  > is turned on.
> 
> Oh yes, netfilter being enabled will cause some performance
> degradation, that is for sure.

Do you think that netfilter being enabled would also cause a decrease in
routing throughput (ie: causing packet loss)?

Cheers!
Jon
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/