From: jamal <hadi@cyberus.ca>
To: Mitch Williams <mitch.a.williams@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>,
john.ronciak@intel.com, jdmason@us.ibm.com, shemminger@osdl.org,
netdev@oss.sgi.com, Robert.Olsson@data.slu.se,
ganesh.venkatesan@intel.com, jesse.brandeburg@intel.com
Subject: Re: RFC: NAPI packet weighting patch
Date: Fri, 03 Jun 2005 15:59:31 -0400 [thread overview]
Message-ID: <1117828771.6071.77.camel@localhost.localdomain> (raw)
In-Reply-To: <Pine.CYG.4.58.0506031202280.3344@mawilli1-desk2.amr.corp.intel.com>
On Fri, 2005-03-06 at 12:28 -0700, Mitch Williams wrote:
>
> On Fri, 3 Jun 2005, David S. Miller wrote:
>
> > From: jamal <hadi@cyberus.ca>
> > Date: Fri, 03 Jun 2005 14:42:30 -0400
> >
> > > When you reduce the weight, the system is spending less time in the
> > > softirq processing packets before softirq yields. If this gives more
> > > opportunity to your app to run, then the performance will go up.
> > > Is this what you are seeing?
> >
> > Jamal, this is my current theory as well, we hit the jiffies
> > check.
>
> Well, I hate to mess up your guys' theories, but the real reason is
> simpler: hardware receive resources, specifically descriptors and
> buffers.
>
> In a typical NAPI polling loop, the driver processes receive packets until
> it either hits the quota or runs out of packets. Then, at the end of the
> loop, it returns all of those now-free receive resources back to the
> hardware.
>
> With a heavy receive load, the hardware will run out of receive
> descriptors in the time it takes the driver/NAPI/stack to process 64
> packets. So it drops them on the floor. And, as we know, dropped packets
> are A Bad Thing.
>
> By reducing the driver weight, we cause the driver to give receive
> resources back to the hardware more often, which prevents dropped packets.
>
> As Ben Greer noticed, increasing the number of descriptors can help with
> this issue. But it really can't eliminate the problem -- once the ring
> is full, it doesn't matter how big it is, it's still full.
>
> In my testing (Dual 2.8GHz Xeon, PCI-X bus, Gigabit network, 10 clients),
> I was able to completely eliminate dropped packets in most cases by
> reducing the driver weight down to about 20.
>
> Now for some speculation:
>
What you said above is unfortunately also speculation ;->
But one that you could validate by putting proper hooks. As an example,
try to restore a descriptor every time you pick one - for an example of
this look at the sb1250 driver.
cheers,
jamal
next prev parent reply other threads:[~2005-06-03 19:59 UTC|newest]
Thread overview: 121+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-06-03 0:11 RFC: NAPI packet weighting patch Ronciak, John
2005-06-03 0:18 ` David S. Miller
2005-06-03 2:32 ` jamal
2005-06-03 17:43 ` Mitch Williams
2005-06-03 18:38 ` David S. Miller
2005-06-03 18:42 ` jamal
2005-06-03 19:01 ` David S. Miller
2005-06-03 19:28 ` Mitch Williams
2005-06-03 19:59 ` jamal [this message]
2005-06-03 20:31 ` David S. Miller
2005-06-03 21:12 ` Jon Mason
2005-06-03 20:22 ` David S. Miller
2005-06-03 20:29 ` David S. Miller
2005-06-03 19:49 ` Michael Chan
2005-06-03 20:59 ` Lennert Buytenhek
2005-06-03 20:35 ` Michael Chan
2005-06-03 22:29 ` jamal
2005-06-04 0:25 ` Michael Chan
2005-06-05 21:36 ` David S. Miller
2005-06-06 6:43 ` David S. Miller
2005-06-03 23:26 ` Lennert Buytenhek
2005-06-05 20:11 ` David S. Miller
2005-06-03 21:07 ` Edgar E Iglesias
2005-06-03 23:30 ` Lennert Buytenhek
2005-06-03 20:30 ` Ben Greear
2005-06-03 19:40 ` jamal
2005-06-03 20:23 ` jamal
2005-06-03 20:28 ` Mitch Williams
-- strict thread matches above, loose matches on Subject: below --
2005-06-07 16:23 Ronciak, John
2005-06-07 20:21 ` David S. Miller
2005-06-08 2:20 ` Jesse Brandeburg
2005-06-08 3:31 ` David S. Miller
2005-06-08 3:43 ` David S. Miller
2005-06-08 13:36 ` jamal
2005-06-09 21:37 ` Jesse Brandeburg
2005-06-09 22:05 ` Stephen Hemminger
2005-06-09 22:12 ` Jesse Brandeburg
2005-06-09 22:21 ` David S. Miller
2005-06-09 22:21 ` jamal
2005-06-09 22:22 ` David S. Miller
2005-06-09 22:20 ` jamal
2005-06-06 20:29 Ronciak, John
2005-06-06 23:55 ` Mitch Williams
2005-06-07 0:08 ` Ben Greear
2005-06-08 1:50 ` Jesse Brandeburg
2005-06-07 4:53 ` Stephen Hemminger
2005-06-07 12:38 ` jamal
2005-06-07 12:06 ` Martin Josefsson
2005-06-07 13:29 ` jamal
2005-06-07 12:36 ` Martin Josefsson
2005-06-07 16:34 ` Robert Olsson
2005-06-07 23:19 ` Rick Jones
2005-06-21 20:37 ` David S. Miller
2005-06-22 7:27 ` Eric Dumazet
2005-06-22 8:42 ` P
2005-06-22 19:37 ` jamal
2005-06-23 8:56 ` P
2005-06-21 20:20 ` David S. Miller
2005-06-21 20:38 ` Rick Jones
2005-06-21 20:55 ` David S. Miller
2005-06-21 21:47 ` Andi Kleen
2005-06-21 22:22 ` Donald Becker
2005-06-21 22:34 ` Andi Kleen
2005-06-22 0:08 ` Donald Becker
2005-06-22 4:44 ` Chris Friesen
2005-06-22 11:31 ` Andi Kleen
2005-06-22 16:23 ` Leonid Grossman
2005-06-22 16:37 ` jamal
2005-06-22 18:00 ` Leonid Grossman
2005-06-22 18:06 ` Andi Kleen
2005-06-22 20:22 ` David S. Miller
2005-06-22 20:35 ` Rick Jones
2005-06-22 20:43 ` David S. Miller
2005-06-22 21:10 ` Andi Kleen
2005-06-22 21:16 ` David S. Miller
2005-06-22 21:53 ` Chris Friesen
2005-06-22 22:11 ` Andi Kleen
2005-06-22 21:38 ` Eric Dumazet
2005-06-22 22:13 ` Eric Dumazet
2005-06-22 22:30 ` David S. Miller
2005-06-22 22:23 ` David S. Miller
2005-06-23 12:14 ` jamal
2005-06-23 17:36 ` David Mosberger
2005-06-22 22:42 ` Leonid Grossman
2005-06-22 23:13 ` Andi Kleen
2005-06-22 23:19 ` David S. Miller
2005-06-22 23:23 ` Andi Kleen
2005-06-22 17:05 ` Andi Kleen
2005-06-06 15:35 Ronciak, John
2005-06-06 19:47 ` David S. Miller
2005-06-03 18:19 Ronciak, John
2005-06-03 18:33 ` Ben Greear
2005-06-03 18:49 ` David S. Miller
2005-06-03 18:59 ` Ben Greear
2005-06-03 19:02 ` David S. Miller
2005-06-03 20:17 ` Robert Olsson
2005-06-03 20:30 ` David S. Miller
2005-06-03 17:40 Ronciak, John
2005-06-03 18:08 ` Robert Olsson
2005-06-02 21:19 Ronciak, John
2005-06-02 21:31 ` Stephen Hemminger
2005-06-02 21:40 ` David S. Miller
2005-06-02 21:51 ` Jon Mason
2005-06-02 22:12 ` David S. Miller
2005-06-02 22:19 ` Jon Mason
2005-06-02 22:15 ` Robert Olsson
2005-05-26 21:36 Mitch Williams
2005-05-27 8:21 ` Robert Olsson
2005-05-27 11:18 ` jamal
2005-05-27 15:50 ` Stephen Hemminger
2005-05-27 20:27 ` Mitch Williams
2005-05-27 21:01 ` Stephen Hemminger
2005-05-28 0:56 ` jamal
2005-05-31 17:35 ` Mitch Williams
2005-05-31 17:40 ` Stephen Hemminger
2005-05-31 17:43 ` Mitch Williams
2005-05-31 22:07 ` Jon Mason
2005-05-31 22:14 ` David S. Miller
2005-05-31 23:28 ` Jon Mason
2005-06-02 12:26 ` jamal
2005-06-02 17:30 ` Stephen Hemminger
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1117828771.6071.77.camel@localhost.localdomain \
--to=hadi@cyberus.ca \
--cc=Robert.Olsson@data.slu.se \
--cc=davem@davemloft.net \
--cc=ganesh.venkatesan@intel.com \
--cc=jdmason@us.ibm.com \
--cc=jesse.brandeburg@intel.com \
--cc=john.ronciak@intel.com \
--cc=mitch.a.williams@intel.com \
--cc=netdev@oss.sgi.com \
--cc=shemminger@osdl.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.