Local bridge performance issue on Linux

[kosztyua]

Hi,

I have the following issue: I wanted to migrate from Windows to Linux, installed a fresh 14.04 x64 ubuntu and configured the latest 9529 softether. Added the physical interface to local bridge and tried performance with a connecting user: abysmal - I mean like 1Mbit or something, compared to the 240Mbit I measure with the same equipment and network and settings, with Windows.
So I started experimenting and created a local bridge to a tap device in softether, and bridged the tap to the physical with bridgeutils. Now I got over 200Mbit as result - still not as expected, but double bridge can lower performance.

Please help me solve this issue. I know SE is originally for Windows, but this is a huge issue.

Thanks,
Andras

EDIT: Could it be MTU issue? In what way is Linux and Windows different?

[GIANT_CRAB]

Hello,

From what I understood, you seem to bridge Softether to Physical Network Adapter and after that, you tried to bridge Softether to a TAP and then bridge the TAP to a physical interface (using bridge-utils), right?

The correct method should be bridging a Softether Virtual Hub to a Physical Network Adapter of your server.

[kosztyua]

Hi,
Thanks for the reply.

Maybe I wasn't clear.
First I bridged the HUB to the Physical in SE -> I experience performance issue
Second I tried bridging the HUB to a TAP in SE and TAP to Physical in Linux -> Almost normal performance

Of course I would go with the direct, first method.

[GIANT_CRAB]

Is SecureNAT disabled for the Hubs you wanted to bridge?

[kosztyua]

Yes of course. In any case, even if I used SecureNAT that would not lower the speed this much, and especially not if the only difference is the operating system.

[thisjun]

How to measure the network throughput?

[kosztyua]

Speedtest.net. I know it is not ideal for academic measurements, but when using the same server, same network, and same speedtest server it can give a good estimation.

But the TAP-bridge and ARP proxy workaround is good, I think I'm going to stick with it, especially because this way I can limit the DHCP broadcast to the TAP-bridge network, while still distributing public IP addresses to the clients.

[PacoBell]

thisjun wrote:
> How to measure the network throughput?

Either use the built-in performance measurement tools:
http://www.softether.org/4-docs/1-manua ... Throughput

or set up iperf3:
http://software.es.net/iperf/