Hi Guys

I have setup an all MikroTik network as below and I am seeing 50% speed loss which doesn't make a whole lot of sense and I was hoping that someone could shed some light on the situation

Router A CHR <-----> Router B 2011RM

This link is an IPIP link and between A <-----> B I get 80Mpbs

Router B <-----> Router C 2011 RM

This link is connected with a PtP link using two LHG radio's and between b <-----> C getting a good 100Mbps

This is where things get a little strange, router C <-----> A gets just 35-40 Mbps

Any thoughts?

Is ptp Full duplex 100mbps. when runing tests are you pushing traffic in both directions.

PtP auto negotiated at 100 full duplex, speed test ran with 1 session receive only

Thanks
Mark

Hi Guys

I have setup an all MikroTik network as below and I am seeing 50% speed loss which doesn't make a whole lot of sense and I was hoping that someone could shed some light on the situation

Router A CHR <-----> Router B 2011RM

This link is an IPIP link and between A <-----> B I get 80Mpbs

Router B <-----> Router C 2011 RM

This link is connected with a PtP link using two LHG radio's and between b <-----> C getting a good 100Mbps

This is where things get a little strange, router C <-----> A gets just 35-40 Mbps

Any thoughts?

TCP over a half duplex wireless link with see a significant reduction in throughput. This can typically be seen when UDP speed tests look pretty good but TCP is not so good. It's really bad when people do multiple hops over half duplex links and wonder why they have such poor results on their speedtest.net speed tests.

PtP auto negotiated at 100 full duplex, speed test ran with 1 session receive only

Thanks
Mark

The ethernet may have negotiated to 100 full duplex but the wireless interface may be half duplex. can you post the model of equipment you are using for your ptp link.

Thanks

Hi Guys

Thanks for the replies - if i am being silly I am more than happy to take it on the chin, this is very new to me.

I have MikroTik LHG at both ends of the PtP link.

The link is showing at 180Mbps tx & rx

Mark

Hi Guys

Thanks for the replies - if i am being silly I am more than happy to take it on the chin, this is very new to me.

I have MikroTik LHG at both ends of the PtP link.

The link is showing at 180Mbps tx & rx

Mark

I hear you. It took me some time myself to understand the dynamics of half duplex and its impacts on TCP performance. As far as your 180Mbps running a speed test from each side of the PTP you see close to 90Mbps on a TCP speed test?

I have been checking through literature and it looks like the LHG will give you around 90 mbps in one direction. As such the 30 - 40 you getting duplex is not bad. here is a link to some tests viewtopic.php?t=111703

Thank you so much for the replies. its great to have a good community to bounce thoughts off.

This still doesn't make sense to me though. I do not see where the full duplex aspect of this comes into play, could someone please clarify, as my speed test between router A and C is still only in one direction

a <-----------> b <-----------> c
80Mbps . 95Mbps
a <----------------------------> c
35 - 40 Mbps

All tests are in one direction only and TCP

Thank you so much for the replies. its great to have a good community to bounce thoughts off.

This still doesn't make sense to me though. I do not see where the full duplex aspect of this comes into play, could someone please clarify, as my speed test between router A and C is still only in one direction

a <-----------> b <-----------> c
80Mbps . 95Mbps
a <----------------------------> c
35 - 40 Mbps

All tests are in one direction only and TCP

TCP is like a high maintenance woman. It does not like to be kept waiting.
Half Duplex is like a CB radio. You can only talk or listen at any one time so this has a huge impact on throughput as its now only able to do one thing at a time. Think of a one lane road with people going each way and having to share that single lane.
Have a peek at the link below to see come cool data. Take note of how latency effects the throughput. And with half duplex sharing a lane that means each side has to wait (latency) to transmit their data. And then you are also expecting to add the sum of each speed test together and get that value on a-c which is still not how TCP works as each hop has its own dynamics. Essentially if the PTP was full duplex you wouldn't see much of an issue at all as the latency would be 0 or very low and your speeds would be as expected.
http://blog.performancevision.com/eng/e ... acket-loss