Network speed slow in openSUSE 11.3?

English Install/Boot/Login
#1

Hi,

Does anyone know how to revert network speeds to those attained with openSUSE11.2-64?

On openSUSE11.2-64 the reported network download speed was,
max : 420 kB/s
ave : ~ 200 kB/s

After upgrade to openSUSE 11.3-64 the figures dropped to about 25% of the previous values.
After upgrade to KDE 4.5 and plasmoid-networkmanagement the initial values did not change but the average value dropped after about 30 seconds to ~ 15 kB/s.

Approach so far:

  1. The motherboard’s (A780GM-LE) build in LAN (Realtek RLT8111DL) was originally detected by openSUSE 11.3 as the Realtek RLT8169 and kernel module r8169 installed.
    This was replaced by the latest module r8168-8.019.00 from Realtek.

  2. After kernel update to Linux 2.6.34.4-0.1-desktop x86_64 the plasmoid-networkmanagement was replace by NetworkManager-kde4. This improved flow with,
    max : 140 kB/s
    ave : ~ 80 kB/s

Any suggestions gratefully accepted.

#2

What kind of network? Windows/Linux

#3

Download/upload confirmed by DSLreports or The National Broadband Plan: Connecting America ?

My guess that it was the driver, ie, the RLT8169 but the RLT8111DL is the right driver for your motherboard, 10/100/1000Mb/s ethernet.

Now my question is whether your router/switch/modem auto-negotiates the 1000mb/s speed or is using the lowest speed?
Or do you have to set the adapter’s speed to 100mb/s. Sometimes the router/switch at 10/100mb/s can’t handle the higher speed.

#4

"Now my question is whether your router/switch/modem auto-negotiates the 1000mb/s speed or is using the lowest speed?
Or do you have to set the adapter’s speed to 100mb/s. Sometimes the router/switch at 10/100mb/s can’t handle the higher speed. "

Everything sets up ok at 100mb/s same as for the old openSUSE 11.2 installation. I havn’t seen the 1000mb/s problem since the initial openSUSE 11.1 with the original kernel.

I’m not having similar problem on openSUSE 11.3-32bit machines on the same network. The speed on the 32 bit machines only used to be about 75% of that achieveable on the 64 bit machine. Now the 32 bit’s are running at least 100% faster than the 64 bit machine.

Thanks for the reply.

#5

"What kind of network? Windows/Linux "

Its a peer to peer network with 2 Windows, 3 Linux machines and 3 printers going to a T-DSL modem for the internet connection. No servers.

Could it be an “ethernet” network?

#6

Well you can open up Wireshark or your router to see if you have collisions and dropped packets over the 11.3-64 connection. The other times I’ve seen extreme dropoff is when the mtu size is corrupt, that should be 1500 or 1492 right? You could turn off IPv6 traffic on the 11.3-64 to see if that helps. Sorry forgot how to do that, as well as dropping the port speed to 100mb/s on the RLT8111DL just to check if its a 100 vs 1000Mb/s issue.

#7

Hi,
For my PC Network thru YAST is much faster than NetworkManager. May be try it…Good Luck…

#8

Generally, local networks including wireless are considered “ethernet” vs. other networks.

Have you checked

 ifconfig -a

for collisions and dropped packets on the slow network port?

#9

NetworkManager is reporting a 100 Mbits/s connection on all machines. Which is the max speed of switch.

'ping’ing the switch from all machines show 0% package loss.

‘sudo ifconfig -a’ does not work due to setup. So is only done as root with internet physically disconnected.
No failures show up.

The OS on the 64 bit machine was an upgrade over openSUSE 11.2. Is it possible the delay is because of remnants of the old OS.

#10

Yes that’s possible. Are you thinking about a clean install? Backup data, hidden files and configuration files like hosts*, sudoers, smb.conf, server data in /var/, etc. Hopefully, /home is on a separate partition from the / (root) partition or you’ll need to back up even more data.

Sometimes the *.rpmnew, *.old , *.bak help if there’s changes not implemented. But even at 100mb/s thru the router should have the same speeds as the 32bit machines.

Any standout error messages in /var/log/messages or /var/log/warn about the interfaces or firewall?

Sorry, but I’m stumped again. You can’t run “ifconfig” as ordinary user or while connected because of the setup?

Are you using knetworkmanager or networkmanager to configure the connection? Or the manual ifup/ifdown?

Your peer-to-peer network connects your PCs to the router thru the router’s internal switch ports or an attached switch?

#11

Hi tararpharazon,

***** Any standout error messages in /var/log/messages or /var/log/warn about the interfaces or firewall?
See results after mini session :-
linux-o2xs:/var/log # cat messages | grep ntpd
Sep 12 18:13:27 linux-o2xs ntpd[3873]: Listening on interface #7 eth0, …#123 Enabled
Sep 12 18:13:27 linux-o2xs ntpd[3873]: new interface(s) found: waking up resolver
Sep 12 18:16:54 linux-o2xs ntpd[3873]: sendto(…) (fd=22): Invalid argument
Sep 12 18:18:00 linux-o2xs ntpd[3873]: sendto(…) (fd=22): Invalid argument
Sep 12 18:18:27 linux-o2xs ntpd[3873]: Deleting interface #7 eth0, …#123, interface stats: received=3, sent=3, dropped=2, active_time=300 secs
Sep 12 18:33:27 linux-o2xs ntpd[3873]: Listening on interface #8 eth0, …#123 Enabled
Sep 12 18:33:27 linux-o2xs ntpd[3873]: new interface(s) found: waking up resolver
Sep 12 18:42:38 linux-o2xs ntpd[3873]: synchronized to …, stratum 3
Sep 12 18:42:38 linux-o2xs ntpd[3873]: kernel time sync status change 6001
Sep 12 18:48:02 linux-o2xs ntpd[3873]: kernel time sync status change 2001
Sep 12 19:08:27 linux-o2xs ntpd[3873]: sendto(…) (fd=22): Invalid argument
Sep 12 19:08:27 linux-o2xs ntpd[3873]: Deleting interface #8 eth0, 192.168.178.21#123, interface stats: received=29, sent=30, dropped=1, active_time=2100 secs
the following warnings only occurred after disconnecting.
linux-o2xs:/var/log # cat warn | grep ntpd
Sep 12 18:16:54 linux-o2xs ntpd[3873]: sendto(…) (fd=22): Invalid argument
Sep 12 18:18:00 linux-o2xs ntpd[3873]: sendto(…) (fd=22): Invalid argument
Sep 12 19:08:27 linux-o2xs ntpd[3873]: sendto(…) (fd=22): Invalid argument

***** Sorry, but I’m stumped again. You can’t run “ifconfig” as ordinary user or while connected because of the setup?
It looks as though a user cannot directly access cmds in /sbin/ where ifconfig is. After making a link to it in /bin/ its now accessable to users.
The results from the last mini session are :-
eth0 Link encap:Ethernet HWaddr ……
inet6 addr: Scope:Link
UP BROADCAST MULTICAST MTU:1500 Metric:1
RX packets:51986 errors:0 dropped:0 overruns:0 frame:0
TX packets:46668 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:69722546 (66.4 Mb) TX bytes:4926287 (4.6 Mb)
Interrupt:25 Base address:0x6000

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:170 errors:0 dropped:0 overruns:0 frame:0
TX packets:170 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:12316 (12.0 Kb) TX bytes:12316 (12.0 Kb)

***** Are you using knetworkmanager or networkmanager to configure the connection? Or the manual ifup/ifdown?
The results above are when plasmoid-networkmanagement is installed.

***** Your peer-to-peer network connects your PCs to the router thru the router’s internal switch ports or an attached switch?
A 5-port switch is before the A-DSL modem. All Linux machines are linked via develo dLANs to the switch, Win machines are hard-wired to the switch.
The dLAN’s reduce transmission speed compared to hard-wired connections but this rate reduction was never been an issue on openSUSE-11.2.
Max download rate on hard-wired connection: 740 kB/s.
The dLANs always indicate a 100 mb/s connection.

Thanks for you replies.

#12

Other than the strange errors for ntpd everything looks as it should.
No collisions or or dropped pakets, no conflicts on IRQ 25.

Another culprit could be plasmoid-networkmanagement as upgraded by 11.3, changelog?

#13

With knetwortmanager installed the speed seems a little better, but not much, wrt plasmoid-networkmanagement.

My assumption, this in because the plasmoid-networkmanagement is collecting more information than knetwortmanager.

If I find a solution I’ll post it here.

Thanks again.

#14

Well this isn’t reporting any of the obvious errors (collisions or dropped), so that’s a help.

A 5-port switch is before the A-DSL modem. All Linux machines are linked via develo dLANs to the switch, Win machines are hard-wired to the switch.

i’d bet you have some kind of problem with the devolo’s. Why it is a problem on one version of Linux and not another, I don’t know, but

  • note that if the network segment between the devolos has low data rate because of collisions (or anything else), this will not show up in the ethernet diagnostics that you gave above (it only shows problems between the computer and the devolo, not from one devolo back to the other devolo and not between that devolo and the 5-post switch; that link will be reported as a healthy, standard, data rate, irrespective of the devolo-to-devolo data rate).
  • the devolos are fairly sensitive to their electrical environment; for example, running one on a muti-way extension is not advised, and too long a cable run between devolos will cause problems too. That said, you are well below the data rate of which they should be capable (which model? which standard?), so it really shouldn’t be a problem at the kind of rate that you can get through your A-DSL box…should… It would be different if you were trying to do local filesharing over them.

The dLANs always indicate a 100 mb/s connection.

As mentioned above, they will do that. It is more or less meaningless for the throughput, or latency, of the total link.

Can you ping the router and post the results, please? I would like to see how fast that runs and how consistent the numbers are.

What about DNS? Does dig show that working quickly, and does it show any speed-up after the first resolution request?

#15

Hi markone,

192.168.178.1 ping statistics —
102 packets transmitted, 102 received, 0% packet loss, time 101161ms
rtt min/avg/max/mdev = 2.396/4.111/12.855/1.963 ms

Query time: 55 msec
;; SERVER: 192.168.178.1#53(192.168.178.1)
;; WHEN: Tue Sep 14 12:17:43 2010
;; MSG SIZE rcvd: 500

#16

Hi markone,

Thanks you’re right!

Long pieces of wire are now stretched to the switch and 640 kB/s is being achieved.

The develo dLANs have been in use since buying SuSE 10.1 in early 2006. Network performance was never a problem before updating to openSuSE 11.3.

No hardware has been changed. Have the sw improvements, superseded the capability of the dLANs?

#17

On 2010-09-13 16:26, keellambert wrote:

> It looks as though a user cannot directly access cmds in /sbin/ where
> ifconfig is. After making a link to it in /bin/ its now accessable to
> users.

That’s intentional. You are using sudo, when you should be using “su -”.

–
Cheers / Saludos,

Carlos E. R.
(from 11.2 x86_64 “Emerald” GM (Minas Tirith))