Speed performance between servers over site to site tunnel

[Mohit_Chauhan]

Hi,

We have site to site (3rd Party) tunnel to Azure from MX100.

The commissioned Internet speed is 250M, however, when we test the speedtest or do throughout test it maximum goes to 160M.

This is not the major concern though - the main issue is that servers on either side of the tunnel show poor network speed. Once this was reported, we did Iperf tests and the number are showing maximum 30Mbps as the bandwidth when using TCP mode. With UDP, it jumps to 150M. Normal speedtest from the server behind the MX shows similar 150M speed as well.

Any suggestions? Also I want to understand why iPerf results are so different from the speedtest, my logic says that speedtest is also using TCP as we use http browser to do the test.

Thanks for going through the above notes. Any suggestions or feedback will be highly appreciated.

[jdsilva]

The default window size in an iPerf TCP test is abysmal. You should be adjusting that to get more accurate real world results. 

[cmr]

TCP traffic requires a lot of processing overhead compared with UDP.  This is because with TCP, both ends need to confirm each part of the conversation.

[PhilipDAth]

Also note that Meraki has crap DES/3DES throughput.  If you are using these ciphers change to AES.

[Mohit_Chauhan]

Hi all,

Thanks for all your contribution into this discussion.

I used TCP with multiple streams which actually showed much better results than a single stream where the network pipe can be tested with more data than just with the limitation of the host hardware.

I also used the below links which also guided me in the testing. Just sharing for all of us, I am sure most of you would have bumped into it at some time.

https://netbeez.net/blog/tcp-window-size/

https://documentation.meraki.com/zGeneral_Administration/Tools_and_Troubleshooting/Troubleshooting_C...

https://www.slashroot.in/iperf-how-test-network-speedperformancebandwidth

Finally thanks all for the help.