Speed loss behind Force 200 PTP link?

HI guys,

Not sure if any of you guys experiecned this weird situation. It is happeneing to multiple sites of our network. 

The network topology is pretty simple:

site A server - Cisco SG300 siwtch - Froce200 PTP (TDD PTP mode) - Cisco SG300 swtich - site B server

The performance of the PTP link is great, over 30 SNR, wireless link test is giving us over 80/80 mbps. IPERF tests are also giving me similar result between site A server and site B server. 

Here comes the interest part, exact the same Interenet speed tests (same target server, test B is conducted right after test A, http://speedtest.telstra.com )on these 2 sites are giving me totally different result. Site B is much slower than site A, results attached. 

It really bothers us beacuse it is happening on muptiple sites now. And customer are using Internet speedtest tool all the time. Our Cisco siwtches are in good shape with plenty of free CPU and memory.

Any suggestion or idea will be very much appericated. 

Ben

Hi Yi,

There is already an open case 112026, where you have asked for the same information and one of my colleague has already send response to you. So I would request you to revert on the same ticket to discuss further.

Regards

Ankur Sharma

Hi,

Yes I did log a ticket with Cambium helpdesk. I'm just trying to see if anyone else has experienced this issue. 

So far, everything I can find leads me to the Force 200 PTP link. Whether it is radio interference or something else I cannot really tell right now. All of our force 200 PTP link are running in clean channels. But we still experience speed loss after Force200 PTP link.

It doesn’t matter if you guys don’t rely on the result of third party speedtest result, neither do we. But please understand the logic and our major concern: let’s say we get 50/50 mbps on third party speed test tool at one server, after a PTP link with 100/100 mbps capacity, we should get at least something close to 50/50 at B end server, right? Exactly same server, same tool, 1 test conducted straight after the other. Right now, we don’t get that. Download is normally fine but upload we lose about a half.

Normally, from my experience with Cambium EPMP series, we only have interference issue when we get a fair bit Error Drop packets. Please see performance statistics as attached. Both uplink and downlink MCS are stabled at 14/15 and link quality/capacity are over 95% consistently.

How does Cambium detect if there is anything on the way blocking signal or not on EPMP radio, like tress or buildings? I do have concert about things are blocking a certain part of signal but EPMP doesn’t have a way to reflect that.

Any idea will be very much apperciated. 

Ben

This sounds a lot like a switch buffer issue.  What else do you have plugged into the Cisco switches?  Do you happen to have any devices with only 100Mbps ports connected to the switches?  What does the topology look like between site A and the internet? 

Hi Jacob,

There are several base stations, a server and other PTP links pluged into Cisco swtich. It's a SG300 switch with 8mb buffer size. Some of them are 100 mbps indeed.

But I did run test bypassing the switch at B end, result was the same. Iperf result between A end sever and B end server was around 60 mbps.

It doesn't look like a siwtch issue to me actually.

Topology:

Gateway 

        |

Csico Swith - A end server (50/50 mbps on internet speed test)

        | 

PTP A end 

        | (70/70 on wireless link test)

PTP B end

        |

B end server (40/20 mbps on internet speed test)

when I tried to run some tests, I noticed the following on throughput chart:

Speed test we got from speedtest. Iperf test are done just between 2 server across the wireless link. To my understanding, these tests are all best effort tests so they will push the data transmission rate to the highest possible. We won’t have any problem if it can always run at the rate of radio wireless link test when a best effort test is requested.

This sounds like buffer bloat like jaccob suggested. Check and see if your switch is tossing pause frames (they will be sent in the direction of the income traffic and not always received from the bloating data stream) Disable flow control and try it again. If you’ve experienced a drastic change in end to end performance, You’ve got bloating headed into one of the other 100 meg devices. If it remains exactly as it then your underlying problem is sitting at a different point. If it is buffer bloat, just add a router to deal with the congestion at that switch and you’ll be back to speed