Outgoing transmission speed unexpected low
In a network I have 3 linux systems. 2 of those are suse systems and the 3rd one is a linux firewall/router (not suse). The suse systems are a 9.2 and a 9.3 system. The outgoing ip/ethernet transmission speed of the 9.3 suse system is only 1.0-2.0 MB/s as reported by scp. While for the incoming traffic it is 6.0MB/s. This 6.0MB/s speed is measured for the other systems (fw <-> 9.2) as well. I wonder what can be reason why the outgoing speed is so low. The network connection has been switched to another network port, but the outgoing speed remains low. Does anyone have an idea what can be the reason for the low outgoing speed? Can this be caused by a bad network card, a hanging network process (although the system is almost idle)?? -- Richard Bos Without a home the journey is endless
In a network I have 3 linux systems. 2 of those are suse systems and the 3rd one is a linux firewall/router (not suse). The suse systems are a 9.2 and a 9.3 system. The outgoing ip/ethernet transmission speed of the 9.3 suse system is only 1.0-2.0 MB/s as reported by scp. While for the incoming traffic it is 6.0MB/s. This 6.0MB/s speed is measured for the other systems (fw <-> 9.2) as well. I wonder what can be reason why the outgoing speed is so low.
The network connection has been switched to another network port, but the outgoing speed remains low.
Does anyone have an idea what can be the reason for the low outgoing speed? Can this be caused by a bad network card, a hanging network process (although the system is almost idle)?? IMHO, that is a meaningless test. I would suggest that you install something
On Friday 05 August 2005 4:52 pm, Richard Bos wrote:
like ttcp or netperf on the 3 systems and run some tests.
I did this several years ago shortly after I got a 10/100 PCMCIA card for my
laptop that was recommended by a friend. I found that the card was giving
me no better than 10Mb.
I returned the card and got a cardbus version that tested close to 100Mb.
These utilities will give you a good test on your LAN. Remember that scp is
bogged down by encryption and some protocol overhead.
--
Jerry Feldman
Op vrijdag 5 augustus 2005 23:04, schreef Jerry Feldman:
IMHO, that is a meaningless test. I would suggest that you install something like ttcp or netperf on the 3 systems and run some tests. I did this several years ago shortly after I got a 10/100 PCMCIA card for my laptop that was recommended by a friend. I found that the card was giving me no better than 10Mb. I returned the card and got a cardbus version that tested close to 100Mb. These utilities will give you a good test on your LAN. Remember that scp is bogged down by encryption and some protocol overhead.
I don't think it is meaningless. scp clearly shows that the up and download speeds are not symmetrical: suse93 -> suse92 6.0MB/s suse93 -> fw 6.0MB/s suse92 -> suse93 1.0MB/s fw -> suse93 1.0MB/s suse92 -> fw 6.0MB/s fw -> suse92 6.0MB/s Now what can be the bottleneck? The network card, the cpu (idling at the moment), well what else. Perhaps the tools you recommend above ttcp and netperf, can tell something about that? -- Richard Bos Without a home the journey is endless
On Fri, 5 Aug 2005 23:28:27 +0200
Richard Bos
Op vrijdag 5 augustus 2005 23:04, schreef Jerry Feldman:
I don't think it is meaningless. scp clearly shows that the up and download speeds are not symmetrical: I would still get a tool like ttcp or netperf, install it on all 3 systems and run a benchmark. Most likely the system has some hardware problem, but scp is not the tool you want to use to diagnose com problems. -- Jerry Feldman
Boston Linux and Unix user group http://www.blu.org PGP key id:C5061EA9 PGP Key fingerprint:053C 73EC 3AC1 5C44 3E14 9245 FB00 3ED5 C506 1EA9
On Friday 05 Aug 2005 21:52, Richard Bos wrote:
In a network I have 3 linux systems. 2 of those are suse systems and the 3rd one is a linux firewall/router (not suse). The suse systems are a 9.2 and a 9.3 system. The outgoing ip/ethernet transmission speed of the 9.3 suse system is only 1.0-2.0 MB/s as reported by scp. While for the incoming traffic it is 6.0MB/s. This 6.0MB/s speed is measured for the other systems (fw <-> 9.2) as well. I wonder what can be reason why the outgoing speed is so low.
I've seen the table in the other posts. Have you tried swapping the cables over and checking the speed again? Are there any potential external influences - one machine on a long line running close to an interference source, for example? Are you getting lots of dropped packets on any machine? Dylan -- "The man who strikes first admits that his ideas have given out." (Chinese Proverb)
Op zaterdag 6 augustus 2005 01:00, schreef Dylan:
I've seen the table in the other posts. Have you tried swapping the cables over and checking the speed again?
That's a good idea.
Are there any potential external influences - one machine on a long line running close to an interference source, for example?
They are all standing next to each other.
Are you getting lots of dropped packets on any machine?
How can I check? -- Richard Bos Without a home the journey is endless
On Saturday 06 Aug 2005 09:25, Richard Bos wrote:
Op zaterdag 6 augustus 2005 01:00, schreef Dylan: <SNIP>
Are there any potential external influences - one machine on a long line running close to an interference source, for example?
They are all standing next to each other.
Fair enough.
Are you getting lots of dropped packets on any machine?
How can I check?
Well, ifconfig will give you a basic count of errors and dropped packets as a starting point. Dylan
Without a home the journey is endless
"The only reason for roaming to to find myself a home"
Op zaterdag 6 augustus 2005 15:29, schreef Dylan:
Are you getting lots of dropped packets on any machine?
How can I check?
Well, ifconfig will give you a basic count of errors and dropped packets as a starting point.
There are some: eth0 Link encap:Ethernet HWaddr 00:10:DC:20:EF:D1 inet addr:192.168.1.4 Bcast:192.168.1.255 Mask:255.255.255.0 inet6 addr: fe80::210:dcff:fe20:efd1/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:119158 errors:2342 dropped:0 overruns:0 frame:0 TX packets:51949 errors:0 dropped:0 overruns:8 carrier:0 collisions:0 txqueuelen:1000 RX bytes:71316809 (68.0 Mb) TX bytes:54031471 (51.5 Mb) Interrupt:11 Base address:0x4000 -- Richard Bos Without a home the journey is endless
On Saturday 06 Aug 2005 15:15, Richard Bos wrote:
Op zaterdag 6 augustus 2005 15:29, schreef Dylan:
Are you getting lots of dropped packets on any machine?
How can I check?
Well, ifconfig will give you a basic count of errors and dropped packets as a starting point.
There are some:
eth0 Link encap:Ethernet HWaddr 00:10:DC:20:EF:D1 inet addr:192.168.1.4 Bcast:192.168.1.255 Mask:255.255.255.0 inet6 addr: fe80::210:dcff:fe20:efd1/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:119158 errors:2342 dropped:0 overruns:0 frame:0 TX packets:51949 errors:0 dropped:0 overruns:8 carrier:0 collisions:0 txqueuelen:1000 RX bytes:71316809 (68.0 Mb) TX bytes:54031471 (51.5 Mb) Interrupt:11 Base address:0x4000
Well, that certainly doesn't look like it would account for the speed difference. Dylan
-- Richard Bos Without a home the journey is endless
-- "The man who strikes first admits that his ideas have given out." (Chinese Proverb)
On Sat, 2005-08-06 at 18:40 +0100, Dylan wrote:
On Saturday 06 Aug 2005 15:15, Richard Bos wrote:
Op zaterdag 6 augustus 2005 15:29, schreef Dylan:
Are you getting lots of dropped packets on any machine?
How can I check?
Well, ifconfig will give you a basic count of errors and dropped packets as a starting point.
There are some:
eth0 Link encap:Ethernet HWaddr 00:10:DC:20:EF:D1 inet addr:192.168.1.4 Bcast:192.168.1.255 Mask:255.255.255.0 inet6 addr: fe80::210:dcff:fe20:efd1/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:119158 errors:2342 dropped:0 overruns:0 frame:0 TX packets:51949 errors:0 dropped:0 overruns:8 carrier:0 collisions:0 txqueuelen:1000 RX bytes:71316809 (68.0 Mb) TX bytes:54031471 (51.5 Mb) Interrupt:11 Base address:0x4000
Well, that certainly doesn't look like it would account for the speed difference.
Dylan
Even though it doesn't look like a lot of errors it is. If possible try using a different nic and see if you get RX errors. Every time there is an error it causes a retransmit of the packets which will slow things down a lot. -- Ken Schneider UNIX since 1989, linux since 1994, SuSE since 1998 "The day Microsoft makes something that doesn't suck is probably the day they start making vacuum cleaners." -Ernst Jan Plugge
----- Original Message -----
From: "Ken Schneider"
On Sat, 2005-08-06 at 18:40 +0100, Dylan wrote:
On Saturday 06 Aug 2005 15:15, Richard Bos wrote:
Op zaterdag 6 augustus 2005 15:29, schreef Dylan:
Are you getting lots of dropped packets on any machine?
How can I check?
Well, ifconfig will give you a basic count of errors and dropped packets as a starting point.
There are some:
eth0 Link encap:Ethernet HWaddr 00:10:DC:20:EF:D1 inet addr:192.168.1.4 Bcast:192.168.1.255 Mask:255.255.255.0 inet6 addr: fe80::210:dcff:fe20:efd1/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:119158 errors:2342 dropped:0 overruns:0 frame:0 TX packets:51949 errors:0 dropped:0 overruns:8 carrier:0 collisions:0 txqueuelen:1000 RX bytes:71316809 (68.0 Mb) TX bytes:54031471 (51.5 Mb) Interrupt:11 Base address:0x4000
Well, that certainly doesn't look like it would account for the speed difference.
Dylan
Even though it doesn't look like a lot of errors it is. If possible try using a different nic and see if you get RX errors. Every time there is an error it causes a retransmit of the packets which will slow things down a lot.
Also try passing the kernel parameter "nolapic" at boot. I had a problem where the nic was flakey/slow and was showing errors. The kernel and the interrupt contoller were not playing nicely together. MSI mobo and Nvidia chipset. The noalpic solved the problem completely.... -- David C. Rankin, J.D., P.E. Rankin * Bertin, PLLC 510 Ochiltree Street Nacogdoches, Texas 75961 (936) 715-9333 www.rankin-bertin.com -- Internal Virus Database is out-of-date. Checked by AVG Anti-Virus. Version: 7.0.323 / Virus Database: 267.9.0/47 - Release Date: 7/16/05
Op zaterdag 6 augustus 2005 22:44, schreef Ken Schneider:
Even though it doesn't look like a lot of errors it is. If possible try using a different nic and see if you get RX errors. Every time there is an error it causes a retransmit of the packets which will slow things down a lot.
I'm going for this options. The system is remote so, it will take some days : (( -- Richard Bos Without a home the journey is endless
Op zaterdag 6 augustus 2005 22:44, schreef Ken Schneider:
Well, that certainly doesn't look like it would account for the speed difference.
Dylan
Even though it doesn't look like a lot of errors it is. If possible try using a different nic and see if you get RX errors. Every time there is an error it causes a retransmit of the packets which will slow things down a lot.
The nic has been replaced and has been running for a day now. This seems to have solved the problem :) -- Richard Bos Without a home the journey is endless
participants (5)
-
David Rankin
-
Dylan
-
Jerry Feldman
-
Ken Schneider
-
Richard Bos