Ticket #268 (assigned defect)

Opened 2 years ago

Last modified 1 year ago

Routing performance problem

Reported by: anonymous Assigned to: wbx (accepted)
Priority: major Milestone: 1.2
Component: kernel Version: trunk
Keywords: Cc:

Description

FreeWRT is a third WRT I tried for this test.

Problem:

FreeWRT (svn 2325, hardware asus wl500g Premium) can route through 54 Mbit/s, OpenWRT (svn 1-2 days ago) on this router can do 44 Mbit/s. Oleg's firmware (http://wl500g.info) can route up to 88 Mbit/s.

Why *WRT's is 2 times slow ?

Oleg said that he uses et.o driver but not b44. I can not find any other substantial difference between *WRT and Oleg's formware.

The test was every time at the same hardware with iperf utility.

I ascked question about in at the http://forum.openwrt.org/viewtopic.php?id=10165 but nobody answeded yet.

Attachments

Change History

04/04/07 10:27:30 changed by wbx

  • owner changed from somebody to wbx.
  • status changed from new to assigned.

Hi,

can you please describe your test setup in detail? This would give us developers the chance to verify and maybe fix the issues. We do not use et.o, because it is proprietary and we cannot use /etc/network/interfaces for the switch configuration. Did you tested 1.0 or trunk? 1.0 has the same drivers and backend code as OpenWrt? whiterussian 0.9, trunk has since two days the same drivers and backend code as OpenWrt? kamikaze.

May be we should discuss this problem on our mailinglist? Because n0-1 has some problems in this area, too.

04/04/07 16:55:16 changed by anonymous

Ok. While I am subscribing to the mail list ...

The test was very simple - classic scheme of the router. At all the tests I did wl500g Premium was a router with the following config:

eth0.0(vlan0) 192.168.1.1/24 eth0.1(vlan1) 10.0.0.1/24 My Note book 10.0.0.10/24, default gateway 10.0.0.1 My Desktop 192.168.1.10/24, ip r add 10.0.0.0/24 via 192.168.1.1 dev eth0 No iptables, no imq, no tc, note book is Intel 1300MHz 256 RAM rtl8139 Desktop is Intel(R) Celeron(R) CPU 1.70GHz 256 RAm rtl8139

Desktop PC and note book have iperf (http://dast.nlanr.net/Projects/Iperf2.0/iperf-2.0.2.tar.gz) On a one side I ran iperf -s on the other sibe iperf -c <IP of the first side>. No metter who was iperf server PC or note book. I get this results:

firmware route speed

1. Oleg's firmware 80-88 Megabits (cool!) (taken from http://wl500g.dyndns.org/#latest)

2. OpenWRT trunk 40-44 Megabits :-( (I do not remember svn version but I svn'ed 2007-03-28)

3. FreeWRT trunk 54 Megabits (svn 2325)

4. Direct through switch 94 Megabits (of course ip addresess was reconfigured PCs was in the same subnet)

At the russian forum Oleg said that et.o was proprietary but already GPL and he uses it becouse et.o is original driver not reverse engineering.

I think the main difference Oleg's firmware from *WRT is a driver for network adapter b44 vs et.o .

08/02/07 13:15:58 changed by tg

  • milestone set to 1.2.

Add/Change #268 (Routing performance problem)