netdev
[Top] [All Lists]

Fw: 2.6.11-mm2 weird ethernet RTTs

To: Stefan Schmidt <s.schmidt@xxxxxxxxxx>
Subject: Fw: 2.6.11-mm2 weird ethernet RTTs
From: Andrew Morton <akpm@xxxxxxxx>
Date: Thu, 17 Mar 2005 11:52:54 -0800
Cc: netdev@xxxxxxxxxxx
Sender: netdev-bounce@xxxxxxxxxxx
Thanks, Stefan.

I'm not aware of any changes which would cause this.  Generally -mm's
networking is the same as Linus's.  Could you test Linus's latest tree?
(-rc1 should be out very soon, so that would be appropriate)


Begin forwarded message:

Date: Thu, 17 Mar 2005 17:05:41 +0100
From: Stefan Schmidt <s.schmidt@xxxxxxxxxx>     
To: akpm@xxxxxxxx
Subject: 2.6.11-mm2 weird ethernet RTTs


Hi Andrew,

since i upgraded to 2.6.11-mm2 i see the following pattern in the
smokeping running on my workstation. The attached pictures show a
behaviour that i also see when pinging (ping -i 0.1) another workstation
on the same LAN and Switch, namely that RTTs (just from and to my
machine, not from any to the other machine) range from 0.7 to 1.8ms,
i.e. rising from 0.7 to 1.8ms in 2-3 seconds and then quickly falling
off to 0.7 in about 1s. (ping output attached)
Trust me: Its not the Switch nor the other workstation.
0.7ms is the minumum RTT when using 1500 byte packets, with 'usual' ping
its like 0.3 to 0.4ms but shows the same behaviour in increasing and
decresing.

Its an Intel Corp. 82547EI Gigabit Ethernet Controller (LOM)
        Subsystem: Asustek Computer, Inc.: Unknown device 80f7
running at 100 Base-TX FD.

Have you got a hint which of the recent networking or scheduler patches
in mm2 could be responsible to this?

best regards,
                Stefan

PS: have replaced IP and Name in the ping output so that you can forward
it to some list if you like.
-- 
"It's just the end-of-MPLS day coming. The second coming of "pure IP" is upon
us."
- Petri Helenius on nanog-ml 

Attachment: frell_last_10800.png
Description: PNG image

Attachment: frell_mini.png
Description: PNG image

Attachment: ping.asc
Description: Text document

<Prev in Thread] Current Thread [Next in Thread>