911G-5HPacD problem with high pps traffic

RouterOS Wireless Networking
1

Hi
We have bought two RB911G-5HPac Device that are using as a PTP Link and we expect to transfer 200Mbps real traffic on it(about 50Kpps)
.
Link Specifications :
CCR1036(ether2)______ RB911G-5HPacD+Dish mANT30 < - - - - - - -2KM FLOS - - - - - > Dish mANT30+ RB911G-5HPacD______AHX2(ether11)
Wireless Link general status such as Signal,CCQ and etc is suitable .
while there is no traffic between CCR and AHx2, ping time is about 0ms .CPU usage status on CCR and AHx2 is completely normal and bandwidth test between two point is great .
Test results from CCR to ahx2

  1. UDP Mode
    • direction=send or receive bandwidth = about 440Mbps
    • direction= both bandwidth = about 200Mbps
    • cpu usage on wireless devices , CCR and AHx2 = completely normal
  2. TCP Mode
    • direction=send or receive bandwidth = about 130Mbps
    • direction= both bandwidth = about 70Mbps
    • cpu usage on wireless devices , CCR and AHx2 = completely normal
    But when we transfer more than 30Mbps real traffic on it, the ping time between two wireless point increases about 10ms or 30 ms that makes latency in the network.
    How can we solve this problem? Is there any relationship between queue-types or other parameters? How much real users traffic (maximum) can we transfer on this device?

Thank you

2

You can use an crystal globe or you can show us the configuration of the wireless. Your choice.

3

:sunglasses: what’s you need ? configure of wireless interface or export from all cfg?

4

Status might help also.

5 



interface=wlan1 radio-name="PTP-AP" mac-address=4C:5E:0C:6F:49:2F ap=yes 
   wds=no bridge=no rx-rate="650Mbps-80MHz/2S/SGI" 
   tx-rate="650Mbps-80MHz/2S/SGI" packets=50928164,55595393 
   bytes=45055148245,60328976519 frames=22368194,27149780 
   frame-bytes=45399978366,60681264193 hw-frames=24277651,27168850 
   hw-frame-bytes=49200929926,61356232901 tx-frames-timed-out=0 uptime=4h5m2s 
   last-activity=0ms signal-strength=-30dBm@HT20-0 signal-to-noise=75dB 
   signal-strength-ch0=-35dBm signal-strength-ch1=-31dBm 
   tx-signal-strength-ch0=-34dBm tx-signal-strength-ch1=-30dBm 
   strength-at-rates=-28dBm@6Mbps 4h4m31s490ms,-30dBm@9Mbps 4h3m47s450ms,-
                  29dBm@12Mbps 4h2m46s790ms,-29dBm@18Mbps 4h2m3s340ms,-
                  30dBm@24Mbps 4h1m1s280ms,-32dBm@36Mbps 4h5s230ms,-32dBm@48Mbps 
                  3h58m40s720ms,-35dBm@54Mbps 3h55m720ms,-30dBm@HT20-0 110ms 
   tx-signal-strength=-28dBm tx-ccq=100% rx-ccq=96% p-throughput=604689 
   nstreme=no framing-mode=none routeros-version="6.27" last-ip=172.16.10.5 
   802.1x-port-enabled=yes management-protection=no compression=no 
   wmm-enabled=yes 
   tx-rate-set="OFDM:6-54 BW:1x-4x SGI:1x-4x HT:0-15 VHTMCS:SS1=0-7,SS2=0-7"
6

Use nv2.

7

Nv2 isn’t optimized for AC… :frowning:

8

Your signal is too high, lower the power to get -50.
I have AC link running on 40MHz channel NV2 with SNR 55dB and is working very well. Link can be fully utilized with 1 TCP connection, max aggregate throughput around 200Mb, latency with full load is 2-3ms, CCQ between 90-99%.

This is with x86 dual core 2133MHz Atom board, R11e-5HacD, Jirous PAR24 Extreme MIMO antenna and 300m distance

9

Really? :laughing:

10

yes…any idea?

11

we test nv2 .. but more throughput possible in 802.11 (bw test resualt)

12

I dont think nv2 is not optimized for AC, people here are using it with success, including me. You have something wrong.
Did you try to change the interface queues and test which one works best with nv2? Did you try to run the bw test between the 911s? What is the throughput when set to 40MHz instead of 80Mhz, did you even try? What is the CCQ when using nv2? The latency increase when using 802.11 is completely normal, you should use nv2…

13

Yes try to lover signals to ~-45dbm , as suggested. using all rates fixed it should be optimal signal strength for best AC wireless performance

also nv2 is working grate on some links on some it doesn’t it depends one environment.

14

-45dBm still seems too high to me. It also depends how high the noise floor is. What is the required SNR for achieving full 256QAM modulation?

15
  • yes i do … best result in multi-queue-ethernet-default
  • yes i try … but when bandwidth go up CPU usage seem to be work as bottleneck
  • yes i try … data rate fixed in 300mbps and throughput in highest level is 220mbps udp and 120 mbps in tcp ( 1 tcp connection ) and ping go around 5-20 ms
  • CCQ is ~90 - 99 when we use NV2
  • when i use NV2 latency increase on speed test (udp or tcp )
16

i try to all rate fix = 18
NV2 isn’t work great for me … signal strength -43 … UDP bw test result isn’t good for me … just 340 mbps and data rate very unstable ( in highest level is 585mbps )

17

Unless something has changed, CCQ doesn’t quite work the same on NV2 as it does on 802.11 and Nstreme. CCQ on NV2 seems to be an average, not a real time display.

I’ll change an AP, see CCQ in the single digits and still get full bandwidth. I’ll check it later and CCQ is 90+ %.

18

after 10 min CCQ go up to +80%

19

Hi all :sunglasses:

thnx 4 all reply

recent change and result :
reduce tx power by all rate fix=18 to ~ 43 signal
NV2 isn’t good for me … TX signal from AP when running bw test very unstable ( -40 > -85 !!! ) why ???

Nstreme is realy great … CCQ is always 100/100 … P troughput ~ 605000 kbps … ping time in 100Mbps traffic = 0-2ms
but in high load station disconnected … log message : Disconnected … too many poll timeouts …

of course best result in 802.11 : udp test 420mbps … tcp test (1 tcp connection ) = 200mbps

Nstreme cfg :
Polling : enable
CSMA : enable
Best fit : 4000

any idea for log message ?!