slow gigabit transfers via samba

[legoman666]

First off let me state that my transfer speeds on the same computer with win xp are rediculously fast (260mb/s) between 2 gigabit enabled computers.

Now with Ubuntu feisty (which i install like 2 days ago, so I know nothing about linux), I'm getting about 90kb/s, which sometimes spikes to 200kb/s for a second or two. Do the math, my speeds over the network decreased by a factor of ~370. This is unacceptble.

I still get the same speeds over http (320kb/s or so) in ubuntu.

my network card is a "RTL8111/8168B PCI Express Gigabit"

with these slow speeds, im not able to stream movies from the other computers on the network (all gigabit, and yes the switch is gigabit obviously) in win xp, im able to stream a hd-dvd or blu ray or other 20000kbps h264 content perfectly fine over the network. Now it took an hour to copy a gigabyte from a windows share.

any clues? if possible, could you be very specific with suggestions? Complete linux newbie here. So far I am liking it, but this is pathetic.

[dmizer]

what's the output of:
ifconfig

what is your setup? are you using a 64bit kernel? have you disabled ipv6? have you tried any other file transfer protocol other than samba (ftp, nfs, scp)?

[legoman666]

eth0      Link encap:Ethernet  HWaddr 00:50:8D:91:0B:FA  
          inet addr:192.168.1.11  Bcast:192.168.1.255  Mask:255.255.255.0
          inet6 addr: fe80::250:8dff:fe91:bfa/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:879101 errors:0 dropped:0 overruns:0 frame:0
          TX packets:604891 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000 
          RX bytes:1247121387 (1.1 GiB)  TX bytes:49079569 (46.8 MiB)
          Interrupt:19 Base address:0xa000 

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:10 errors:0 dropped:0 overruns:0 frame:0
          TX packets:10 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0 
          RX bytes:668 (668.0 b)  TX bytes:668 (668.0 b)

im using the 32bit kernel. No, I have not disabled ipv6 and do not know how. And no, I have not tried any other transfer method yet, nor do I know how with linux. Still working on it.

[kelvinchow]

Originally Posted by legoman666 eth0 Link encap:Ethernet HWaddr 00:50:8D:91:0B:FA inet addr:192.168.1.11 Bcast:192.168.1.255 Mask:255.255.255.0 inet6 addr: fe80::250:8dff:fe91:bfa/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:879101 errors:0 dropped:0 overruns:0 frame:0 TX packets:604891 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:1247121387 (1.1 GiB) TX bytes:49079569 (46.8 MiB) Interrupt:19 Base address:0xa000 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:10 errors:0 dropped:0 overruns:0 frame:0 TX packets:10 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:668 (668.0 b) TX bytes:668 (668.0 b) im using the 32bit kernel. No, I have not disabled ipv6 and do not know how. And no, I have not tried any other transfer method yet, nor do I know how with linux. Still working on it.

Your MTU of eth0 is 1500 only , adjust MTU up to 9000 , that will be fine

[netztier]

Originally Posted by kelvinchow Your MTU of eth0 is 1500 only , adjust MTU up to 9000 , that will be fine

This is only advisable if all NICs participating in that nework support Jumbo Frames and also the (wired) switch does. All systems on that subnet should be running the PathMTU Discovery feature, and the router giving access to the Internet from such an MTU 9000 enabled network must be able to respond with an ICMP message type 3 code 4 ("Fragmentation needed, but DF bit set") .

Combining MTU 9000 with a LAN that has a MTU 1500 wireless segment bridged to it (as it commonly happens with broadband wireless routers) will lead to connectivity problems.

Be wary with fiddling with the MTU...

best regards

Marc

[=JoNaZ=]

i have the exact same problem... bought a new motherboard.. gigabyte p32-ds3 i think its the same lan controller... solution anyone?

[chrisp21]

I haven't gotten into the world of gigabit yet, but when building a linux machine I always buy a pci ethernet adapter that I know works well with Linux and turn off the integrated motherboard adapter. Some drivers just don't get the most of the hardware because the developer doesn't know much about the actual hardware - because there are no specs. A majority of the integrated adapters tend to be more obscure hardware and don't work as well. It sounds to me like this is the situation you're running into.

You may be able to find a better driver, but my guess is that you'll need to get a gigabit card that is designed for linux (there are quite a few). If you're liking Linux (other than this issue) than it's probably worth it.

[un_dave]

I have a similar motherboard, p35-ds3p, and i'm experiencing exactly the same problem.

I setup samba shares, so other networked windows/linux/xbox boxes could access the media on this machine, but interestingly i get performance issues with all but one.

Connecting to my debian xbox, and transferring via ssh2, i get speeds of:
1.3MB/s which is reasonable, yet if i connect via samba, i get only a few kb/s.

None of my other systems have any issues with there samba shares.

Although the chipset supports gigabit ethernet, my switches only support 100mbit.

ifconfig gives:

eth0 Link encap:Ethernet HWaddr 00:1A:4D:4B:E1:6F
inet addr:10.1.1.103 Bcast:10.1.1.255 Mask:255.255.255.0
inet6 addr: fe80::21a:4dff:fe4b:e16f/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:3685 errors:0 dropped:0 overruns:0 frame:0
TX packets:3461 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:3114339 (2.9 MiB) TX bytes:892359 (871.4 KiB)
Interrupt:17 Base address:0xe000


Anyone have any ideas?

[reckless2k2]

what are the results when you run the following command

sudo ethtool eth0 | grep -i speed

it should say

Speed: 1000Mb/s

????

[un_dave]

dave@hades:~$ sudo ethtool eth0 | grep -i speed
Speed: 100Mb/s

That makes sense though, because i'm connected to a 100Mb/s switch, so it won't be running at 1000Mb/s.

Another interesting point, I just tried pulling data from a windows xp machine's share, and got a good transfer speed, and writing into a share on the same computer also gives a good speed.

Does that help things? Or just confuse them?