A5 Traffic Shaping Rates: Rates in PHY or MAC?


#1

So running into an interesting issue, where traffic shaping plans seem to be only giving ~ 50% of the speed. Does anyone know if the traffic shapping Mbps is the phy limit or the mac limit?

Ex: IF I want to provision a real world TCP speed test (assuming all other things like uplink /ethernet/routing performance exceeds these limits) a 50 Down and 50 Up so I need to configure the Mbps rating based on the phy *0.5 * mac efficency? Or is the Traffic shaping supose to be the mac rate?

Tim


#2

I am pretty sure you should be getting the BW Limiting on the MAC rate, but I personally don’t use the BW Limiting in the A5 + C5 stuff. (To be honest, we don’t trust any radio to do much besides be a stupid bridge, seems hard enough for them to do that.)

There was another guy who was having issues similar to yours: A5c Performance Issues with 100+mbps Plans

I would say start with my last comments, “How are you doing the testing?” That will help show where the problem(s) might be…


#3

@Amplex, what firmware version are you running?

Older firmware had terrible traffic shaping, and would behave like you described. Recent firmware, however, seems to be close to right on for me.


#4

2.5.1

I’m going to lab it up shortly and see what I find.


#5

So after doing some testing today I can say there is definitely an interesting issue on the downlink side. The uplink side (client to AP) rate limits correctly but the downlink side is consistently ranging from 65 to 45% of the listed speed in the interface.

So I kept playing around with settings and found that if I didn’t set the “commit” to the same speed as the “max” then the rate limiting on the downlink side saw a range of ~ 65% to 45% of the listed speed.

To make all things equal I stated with a non-capped plan and ran a test using both our in-house speedtest.net server with my laptop connected to the C5c, and then using a RB4011 Mikrotik connected to a CHR bandwidth test server.

I set the plan in traffic shaping to default 1000/100 and received speeds of ~ 232 down and ~ 116 up over a period of 1 minute.

Setting the traffic shaping to 55 max down 10 down commit, 55 uplink max, 10 uplink commit resulted in 29.8 download and 55.1 upload.

Changing this to 55 max down 55 commit, 55 uplink max, 55 uplink commit resulted in 55 down and 55 up.

@Mimosa any chance you guys can investigate this and patch it?


#6

Calling: @DustinS & @DavidD & @David

Out of my ball field. Figure one of you might know something.


#7

What was the speed when you have no traffic shaping in place?

So you have an A5 in house with no other clients connected to it?


#8

Yes, this is a brand new A5c with only 1 client (my c5c) connected in-house on clear RF channel without any rate limiting (the default 1000/1000 maxium) ~ 232 down and ~ 116 up over a period of 1 minute.

It clearly doing higher then any of the speeds without the rate limiting applied. The speed issue only appeared when I set the “commit” rate on the downlink side to a lower number then the maximum. Ex: 55 download max with a commit of 10 resulted in speeds of ~ 29 down.

I just re-re-ran those tests and exact same result, the downlink commit being set lower then the max speeds (ex: 55 max 10 commit) resulted in same speed reduction. As does other combinations (105/20, 35/5)

This does not appear to be true on uplink side, the commit rate did not seem to effect the speed negatively.

Interest pattern, so far, the rate seems to scale as commit goes up roughly to be the average of the (max + commit ) / 2. Ex: 55max/20commit, I got a rate of ~38.8 UDP. Setting it to 55max/5commit I got a rate of 30.5.

Tim


#9

Very interesting… As I said. I don’t do much with the Bandwidth limiting, but I have messaged all the @Mimosa people I know on the forum.

I know DustinS is on the FaceBook stuff pretty frequently.

You might just want to get ahold of the Mimosa support guy directly (they have a email you can reach them on) mimosa.co/support, or via the “Chat with us” located on the lower right of most Mimosa web pages (including the forum).

If you do solve the issue outside of the forum, please report back to the community. It will probably help with solving other people’s issues…