On 2007/10/02 , at 18:47, Stephen Hemminger wrote:
On Tue, 2 Oct 2007 09:25:34 -0700
[EMAIL PROTECTED] (Larry McVoy) wrote:
If the server side is the source of the data, i.e, it's transfer
is a
write loop, then I get the bad behaviour.
...
So is this a bug or intentional?
For whatever it
On Mon, 15 Oct 2007 14:40:25 +0200
Daniel Schaffrath [EMAIL PROTECTED] wrote:
On 2007/10/02 , at 18:47, Stephen Hemminger wrote:
On Tue, 2 Oct 2007 09:25:34 -0700
[EMAIL PROTECTED] (Larry McVoy) wrote:
If the server side is the source of the data, i.e, it's transfer
is a
write
Tangential aside:
On Tue, 02 Oct 2007, Rick Jones wrote:
*) depending on the quantity of CPU around, and the type of test one is
running,
results can be better/worse depending on the CPU to which you bind the
application. Latency tends to be best when running on the same core as takes
From: [EMAIL PROTECTED] (Larry McVoy)
Date: Tue, 2 Oct 2007 15:36:44 -0700
On Tue, Oct 02, 2007 at 03:32:16PM -0700, David Miller wrote:
I'm starting to have a theory about what the bad case might
be.
A strong sender going to an even stronger receiver which can
pull out packets into
A few notes to the discussion. I've seen one e1000 bug that ended up being
a crappy AMD pre-opteron SMP chipset with a totally useless PCI bus
implementation, which limited performance quite a bit-totally depending on
what you plugged in and in which slot. 10e milk-and-bread-store
32/33 gige
On Tue, Oct 02, 2007 at 02:21:32PM -0700, Larry McVoy wrote:
More data, sky2 works fine (really really fine, like 79MB/sec) between
Linux dylan.bitmover.com 2.6.18.1 #5 SMP Mon Oct 23 17:36:00 PDT 2006 i686
Linux steele 2.6.20-16-generic #2 SMP Sun Sep 23 18:31:23 UTC 2007 x86_64
So this is
On Wed, Oct 03, 2007 at 02:23:58PM -0700, Larry McVoy wrote:
A few notes to the discussion. I've seen one e1000 bug that ended up being
a crappy AMD pre-opteron SMP chipset with a totally useless PCI bus
implementation, which limited performance quite a bit-totally depending on
what you
Larry McVoy [EMAIL PROTECTED] wrote:
One of my clients also has gigabit so I played around with just that
one and it (itanium running hpux w/ broadcom gigabit) can push the load
as well. One weird thing is that it is dependent on the direction the
data is flowing. If the hp is sending then
Larry McVoy wrote:
A short summary is can someone please post a test program that sources
and sinks data at the wire speed? because apparently I'm too old and
clueless to write such a thing.
Here's a simple reference tcp source/sink that's I've used for years.
For example, on a couple
On Tue, Oct 02, 2007 at 06:52:54PM +0800, Herbert Xu wrote:
One of my clients also has gigabit so I played around with just that
one and it (itanium running hpux w/ broadcom gigabit) can push the load
as well. One weird thing is that it is dependent on the direction the
data is flowing.
Interesting data point. My test case is like this:
server
bind
listen
while (newsock = accept...)
transfer()
client
connect
transfer
If the server side is the source of the data, i.e, it's transfer is a
write loop, then I get the bad
If the server side is the source of the data, i.e, it's transfer is a
write loop, then I get the bad behaviour.
...
So is this a bug or intentional?
For whatever it is worth, I believed that we used to get better performance
from the same hardware. My guess is that it changed somewhere
On Tue, 2 Oct 2007, Larry McVoy wrote:
Interesting data point. My test case is like this:
server
bind
listen
while (newsock = accept...)
transfer()
client
connect
transfer
If the server side is the source of the data, i.e, it's
On Tue, 2 Oct 2007 09:25:34 -0700
[EMAIL PROTECTED] (Larry McVoy) wrote:
If the server side is the source of the data, i.e, it's transfer is a
write loop, then I get the bad behaviour.
...
So is this a bug or intentional?
For whatever it is worth, I believed that we used to get
Isn't this something so straightforward that you would have tests for it?
This is the basic FTP server loop, doesn't someone have a big machine with
10gig cards and test that sending/recving data doesn't regress?
Sounds like a bug to me, modulo the above caveat of making sure that it's
not
On Tue, Oct 02, 2007 at 09:47:26AM -0700, Stephen Hemminger wrote:
On Tue, 2 Oct 2007 09:25:34 -0700
[EMAIL PROTECTED] (Larry McVoy) wrote:
If the server side is the source of the data, i.e, it's transfer is a
write loop, then I get the bad behaviour.
...
So is this a bug or
Larry McVoy wrote:
Interesting data point. My test case is like this:
server
bind
listen
while (newsock = accept...)
transfer()
client
connect
transfer
If the server side is the source of the data, i.e, it's transfer is a
write loop,
I have a more complex configuration application, but I don't see this
problem in my testing. Using e1000 nics and modern hardware
I'm using a similar setup, what kernel are you using?
I am purposefully setting the socket send/rx buffers, as well has
twiddling with the tcp and netdev
On Tue, 2 Oct 2007 09:49:52 -0700
[EMAIL PROTECTED] (Larry McVoy) wrote:
On Tue, Oct 02, 2007 at 09:47:26AM -0700, Stephen Hemminger wrote:
On Tue, 2 Oct 2007 09:25:34 -0700
[EMAIL PROTECTED] (Larry McVoy) wrote:
If the server side is the source of the data, i.e, it's transfer is a
Larry McVoy wrote:
A short summary is can someone please post a test program that sources
and sinks data at the wire speed? because apparently I'm too old and
clueless to write such a thing.
http://www.netperf.org/svn/netperf2/trunk/
:)
WRT the different speeds in each direction talking
Larry McVoy wrote:
I have a more complex configuration application, but I don't see this
problem in my testing. Using e1000 nics and modern hardware
I'm using a similar setup, what kernel are you using?
I'm currently on 2.6.20, and have also tried 10gbe nics on 2.6.23 with
good
On Tue, Oct 02, 2007 at 10:14:11AM -0700, Rick Jones wrote:
Larry McVoy wrote:
A short summary is can someone please post a test program that sources
and sinks data at the wire speed? because apparently I'm too old and
clueless to write such a thing.
WRT the different speeds in each
I'm currently on 2.6.20, and have also tried 10gbe nics on 2.6.23 with
My guess is that it is a bug in the debian 2.6.18 kernel.
Have you tried something like ttcp, iperf, or even regular ftp?
Yeah, I've factored out the code since BitKeeper, my test program,
and John's test program all
On Tue, 2 Oct 2007 10:21:55 -0700
[EMAIL PROTECTED] (Larry McVoy) wrote:
I'm currently on 2.6.20, and have also tried 10gbe nics on 2.6.23 with
My guess is that it is a bug in the debian 2.6.18 kernel.
Have you tried something like ttcp, iperf, or even regular ftp?
Yeah, I've factored
has anyone already asked whether link-layer flow-control is enabled?
rick jones
-
To unsubscribe from this list: send the line unsubscribe netdev in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Larry McVoy wrote:
On Tue, Oct 02, 2007 at 06:52:54PM +0800, Herbert Xu wrote:
One of my clients also has gigabit so I played around with just that
one and it (itanium running hpux w/ broadcom gigabit) can push the load
as well. One weird thing is that it is dependent on the direction the
data
Make sure you don't have slab debugging turned on. It kills performance.
It's a stock debian kernel, so unless they turn it on it's off.
--
---
Larry McVoylm at bitmover.com http://www.bitkeeper.com
-
To unsubscribe from this list: send the line unsubscribe netdev in
On Tue, Oct 02, 2007 at 11:01:47AM -0700, Rick Jones wrote:
has anyone already asked whether link-layer flow-control is enabled?
I doubt it, the same test works fine in one direction and poorly in the other.
Wouldn't the flow control squelch either way?
--
---
Larry McVoylm at
Looks like you have TSO enabled. Does it behave differently if it's
disabled?
It cranks the interrupts/sec up to 8K instead of 5K. No difference in
performance other than that.
I think Rick Jones is on to something with the HP ack avoidance.
I sincerely doubt it. I'm only using the
On Tue, 2 Oct 2007, Larry McVoy wrote:
tcpdump is a good idea, take a look at this. The window starts out
at 46 and never opens up in my test case, but in the rsh case it
starts out the same but does open up. Ideas?
I don't think that's an issue, since you only send one way. The window
On Tue, 2 Oct 2007, Larry McVoy wrote:
No HP in the mix. It's got nothing to do with hp, nor to do with rsh, it
has everything to do with the direction the data is flowing.
Can you tcpdump both cases and send snippets (both of steady-state, and
the initial connect)?
More data, we've conclusively eliminated the card / cpu from the mix.
We've got 2 ia64 boxes with e1000 interfaces. One box is running
linux 2.6.12 and the other is running hpux 11.
I made sure the linux one was running at gigabit and reran the tests
from the linux/ia64 = hp/ia64. Same results,
Larry McVoy wrote:
On Tue, Oct 02, 2007 at 11:01:47AM -0700, Rick Jones wrote:
has anyone already asked whether link-layer flow-control is enabled?
I doubt it, the same test works fine in one direction and poorly in the other.
Wouldn't the flow control squelch either way?
While I am often
Larry McVoy wrote:
More data, we've conclusively eliminated the card / cpu from the mix.
We've got 2 ia64 boxes with e1000 interfaces. One box is running
linux 2.6.12 and the other is running hpux 11.
I made sure the linux one was running at gigabit and reran the tests
from the linux/ia64 =
I also would have expected more ACK's from the HP box. It's been a long
time since I did TCP, but I thought the rule was still that you were
supposed to ACK at least every other full frame - but the HP box is acking
roughly every 16K (and it's *not* always at TSO boundaries: the earlier
ACK's
I think I'm still missing some basic data here (probably because this
thread did not originate on netdev). Let me try to nail down some of
the basics. You have a linux ia64 box (running 2.6.12 or 2.6.18?) that
sends slowly, and receives faster, but not quite a 1 Gbps? And this is
true
From: Linus Torvalds [EMAIL PROTECTED]
Date: Tue, 2 Oct 2007 12:29:50 -0700 (PDT)
On Tue, 2 Oct 2007, Larry McVoy wrote:
No HP in the mix. It's got nothing to do with hp, nor to do with rsh, it
has everything to do with the direction the data is flowing.
Can you tcpdump both cases
From: Linus Torvalds [EMAIL PROTECTED]
Date: Tue, 2 Oct 2007 12:27:53 -0700 (PDT)
We see a single packet containing 16060 bytes, which seems to be because
of TSO on the sending side (you did your tcpdump on the sender, no?), so
it will actually be broken up into 11 1460-byte regular frames
Alternatively, take your favorite test programs, such as John's,
and make a second pair that reverses the direction the data is
sent. So one pair is server sends, the other is server receives,
try both. That's where we started, BitKeeper, my stripped down test,
and John's test all exhibit the
It would be really great to see numbers with a more recent kernel
than 2.6.18
FWIW Debian has binaries for 2.6.21 in testing and for 2.6.22 in
unstable so it should be very easy for Larry to try at least those.
- R.
-
To unsubscribe from this list: send the line unsubscribe netdev in
the
From: [EMAIL PROTECTED] (Larry McVoy)
Date: Tue, 2 Oct 2007 09:48:58 -0700
Isn't this something so straightforward that you would have tests for it?
This is the basic FTP server loop, doesn't someone have a big machine with
10gig cards and test that sending/recving data doesn't regress?
We fixed a lot of bugs in TSO last year.
It would be really great to see numbers with a more recent kernel
than 2.6.18
More data, sky2 works fine (really really fine, like 79MB/sec) between
Linux dylan.bitmover.com 2.6.18.1 #5 SMP Mon Oct 23 17:36:00 PDT 2006 i686
Linux steele
On Tue, Oct 02, 2007 at 02:16:56PM -0700, David Miller wrote:
We absolutely depend upon people like you to report when there are
anomalies like this. It's the only thing that scales.
Well cool, finally doing something useful :)
Is this issue no test setup? Because this does seem like
From: [EMAIL PROTECTED] (Larry McVoy)
Date: Tue, 2 Oct 2007 14:26:08 -0700
And note that sky2 doesn't have this problem. Does the broadcom do TSO?
And sky2 not? I noticed a much higher CPU load for sky2.
Yes the broadcoms (the revisions I have) do TSO and it is enabled
on both sides.
Which
From: [EMAIL PROTECTED] (Larry McVoy)
Date: Tue, 2 Oct 2007 11:40:32 -0700
I doubt it, the same test works fine in one direction and poorly in the other.
Wouldn't the flow control squelch either way?
HW controls for these things are typically:
1) Generates flow control flames
2) Listens for
On Tue, 2 Oct 2007, Wayne Scott wrote:
The slow set was done like this:
on ia64: netcat -l -p /dev/null
on work: netcat ia64 /dev/zero
That sounds wrong. Larry claims the slow case is when the side that did
accept() does the sending, the above has the listener just
David Miller wrote:
From: [EMAIL PROTECTED] (Larry McVoy)
Date: Tue, 2 Oct 2007 14:26:08 -0700
And note that sky2 doesn't have this problem. Does the broadcom do TSO?
And sky2 not? I noticed a much higher CPU load for sky2.
Yes the broadcoms (the revisions I have) do TSO and it is
From: Rick Jones [EMAIL PROTECTED]
Date: Tue, 02 Oct 2007 15:17:35 -0700
Stranger still, with a mix of a 2.6.23-rc5ish kernel and a net-2.6.24 one
(pulled oh middle of last week?) I get link-rate and I see no asymmetry
between
TCP_STREAM and TCP_MAERTS over an e1000 link with no switch or
On Tue, Oct 02, 2007 at 03:32:16PM -0700, David Miller wrote:
I'm starting to have a theory about what the bad case might
be.
A strong sender going to an even stronger receiver which can
pull out packets into the process as fast as they arrive.
This might be part of what keeps the receive
Larry McVoy wrote:
On Tue, Oct 02, 2007 at 03:32:16PM -0700, David Miller wrote:
I'm starting to have a theory about what the bad case might
be.
A strong sender going to an even stronger receiver which can
pull out packets into the process as fast as they arrive.
This might be part of what
On Sat, Sep 29, 2007 at 11:02:32AM -0700, Linus Torvalds wrote:
On Sat, 29 Sep 2007, Larry McVoy wrote:
I haven't kept up on switch technology but in the past they were much
better than you are thinking. The Kalpana switch that I had modified
to support vlans (invented by yours truly), did
On Mon, 1 Oct 2007, Larry McVoy wrote:
but the client looks like
connect(3, {sa_family=AF_INET, sin_port=htons(31235),
sin_addr=inet_addr(10.3.9.1)}, 16) = 0
read(3, \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0..., 1048576)
= 2896
read(3,
On Mon, Oct 01, 2007 at 07:14:37PM -0700, Linus Torvalds wrote:
On Mon, 1 Oct 2007, Larry McVoy wrote:
but the client looks like
connect(3, {sa_family=AF_INET, sin_port=htons(31235),
sin_addr=inet_addr(10.3.9.1)}, 16) = 0
read(3,
From: [EMAIL PROTECTED] (Larry McVoy)
Date: Mon, 1 Oct 2007 19:20:59 -0700
A short summary is can someone please post a test program that sources
and sinks data at the wire speed? because apparently I'm too old and
clueless to write such a thing.
You're not showing us your test program so
On Mon, Oct 01, 2007 at 08:50:50PM -0700, David Miller wrote:
From: [EMAIL PROTECTED] (Larry McVoy)
Date: Mon, 1 Oct 2007 19:20:59 -0700
A short summary is can someone please post a test program that sources
and sinks data at the wire speed? because apparently I'm too old and
clueless
55 matches
Mail list logo