Hi,

Currently,I have a PtP link for distance 10Km at 5Ghz.
Both side Using Rb411A+CM9. One as bridge and other as station.
Nstreme enable, Enable Polling, disable CSMA, framer policy dynamic size.

The signal strength is -69/-70 and TX/RX CCQ around 85-100% and I lock it at 36Mbps.
Signal to noise is around 22-27

I can get stable 18-21Mbps when i'm doing bandwidth test between wireless IP address.
And here is the ping result :

[admin@btv-jrk] > ping 10.1.0.9
10.1.0.9 64 byte ping: ttl=64 time=12 ms
10.1.0.9 64 byte ping: ttl=64 time=10 ms
10.1.0.9 64 byte ping: ttl=64 time=9 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=9 ms
10.1.0.9 64 byte ping: ttl=64 time=9 ms
10.1.0.9 64 byte ping: ttl=64 time=9 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=10 ms
10.1.0.9 64 byte ping: ttl=64 time=19 ms
10.1.0.9 64 byte ping: ttl=64 time=49 ms
10.1.0.9 64 byte ping: ttl=64 time=9 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=10 ms
10.1.0.9 64 byte ping: ttl=64 time=19 ms
15 packets transmitted, 15 packets received, 0% packet loss
round-trip min/avg/max = 1/11.8/49 ms

I'm quite satisfied with the bandwidth test result but not with the ping time.
Already try to change to some freq but the result is the same.

I also try to turn off the Nstreme, the ping time become very good, but I only get half of Mbps when compared with Nstreme enable.

[admin@btv-jrk] > ping 10.1.0.9
10.1.0.9 64 byte ping: ttl=64 time=2 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
10.1.0.9 64 byte ping: ttl=64 time=2 ms
10.1.0.9 64 byte ping: ttl=64 time=1 ms
13 packets transmitted, 13 packets received, 0% packet loss
round-trip min/avg/max = 1/1.1/2 ms

Is there anyway to improve the ping time with Nstreme enable? What setting and condition affect the ping time on wireless mikrotik device?

Any help will be appreciated.

Thank you

Aldhy

You can't have both.

Nstream improves throughput by combining multiple small packets into a single larger one. Thus pings are delayed while waiting to be combined.

You saw this when you turned it off.

So what do you want, more throughput or more consistent pings? The choice is yours.

Tom

BTW I found that 'best fit' and '3200' works best for me when using nstream.

Hi Tom,

Thank you for your reply

I realize that I can not have both advantage. And I choose to enable the Nstreme

Personally I would like to see some modifications to modify N-stream to give a more consistent ping albeit higher. I'd rather see 10ms all the time then fluctuation in pings.

Personally I would like to see some modifications to modify N-stream to give a more consistent ping albeit higher. I'd rather see 10ms all the time then fluctuation in pings.

yes the fluctuation in ping times are the main problem in nstreme links.

I suggest you to read Toms reply again. This is how Nstreme works. It waits to combine packets into bigger ones. This is specifially made so. If we would stop doing this (to save your pings) there would be absolutely no use to Nstreme.

try use wireless-test package.
I believe it's been discussed somewhere....

Hope this help.

Normis,

My suggestion is to formulate something within Nstream to increase the latency but keep it consistant. I understand how Nstream works and why the latency spikes are evident. I would rather see consistantly higher pings then erratic ones.

... and what about - framer policy=none?
i.e. if u want the polling only, anyone tried this? thanks.

Bay good Dish Antennas

jwcn,
i think that suggestion to normis is kinda tricky, as the only way you can have a consistent ping (theoretically) with nstreme would be to have a constant amount of traffic passing through the interface...as it has to wait for each frame to be filled. thats also assuming its a ptp link. it might improve with framer-policy=best-fit but i guess that's up for debate since i don't code for ROS

Tom's explanation is the best i believe.