sxt-lite5 mt6.7 mimo speed problem

[gringoZ]

hi

i have problem with throughput/speed at 520m link working on sxt-lite5/mt6.7 802.11n (no nstreme, no nv2).

1. speedtest.net mini, instaled at server from laptop-2 i had max 40/40 mbit
2. speedtest.net mini, instaled at server from laptop-1 i had max 90/90 mbit
3. bandwidth test (10 connection) from mt-pc-2 to mt-pc-1 —> 80/80 mbit
4. bandwidth test (1 connection) from mt-pc-2 to mt-pc-1 —> 31/25 mbit
5. real internet traffic from sxt-lite5-2 to sxt-lite5-1 —> 40/25 mbit

why test 1 result is different from 2?
why test 3 result is different from 4?

i try bridge/vpls/station bridge/pseudo/wds/nstreme/nv2 and change frequency, and i cant improve throughput of this link…

how i can improve this link? what is bottleneck (sxt cpu, latency)?

[plisken]

Reduce the power of the sxt so that your signal is -60dB

It can be that the signal is to big and disturbe itselfs.

By wireless HT select chain 0 and 1

[gringoZ]

@plisken

By wireless HT select chain 0 and 1

look at picture description, i had 300/300Mbit than i must have chain 0 and 1 included

@plisken

Reduce the power of the sxt so that your signal is -60dB
It can be that the signal is to big and disturbe itselfs.

i tried set at 1, 2, 3, 4, 5dBm and throughput not improved but ccq is worse, jumps 88-100% and tx/rx jumps 216-300Mbit

[plisken]

Have you a line of sight or do you have obstacles between the link

[gringoZ]

clear LOS, distance 250m
sxt-lite5-1 at 10 floor of tower block, sxt-lite5-2 at 8 floor of tower block, they are propely tilted up/down

[InoX]

Speedtest mini is not an reliable speedtest.
Btest same as speedtest.
Why don’t try an torrent?

[gringoZ]

@InoX

Speedtest mini is not an reliable speedtest.
Btest same as speedtest.
Why don’t try an torrent?

i understand that speedtest from ookla isnt prefect. but almost everyone use it.

how can i improve throughput of my link?

[plisken]

bandwidth test in tools

[normis]

@plisken

bandwidth test in tools

the bandwidth test should be run from other router behind the wireless link, because it stresses the CPU, and will affect the results.

[gringoZ]

i found http://smutz.us/techtips/NetworkLatency.html how it is that I have ping from laptop-2 to server:

ping server (10.0.0.40) -c100
PING server (10.0.0.40) 56(84) bytes of data.
64 bytes from server (10.0.0.40): icmp_req=1 ttl=64 time=8.50 ms
64 bytes from server (10.0.0.40): icmp_req=2 ttl=64 time=5.52 ms
64 bytes from server (10.0.0.40): icmp_req=3 ttl=64 time=2.48 ms
...
64 bytes from server (10.0.0.40): icmp_req=98 ttl=64 time=0.764 ms
64 bytes from server (10.0.0.40): icmp_req=99 ttl=64 time=2.67 ms
64 bytes from server (10.0.0.40): icmp_req=100 ttl=64 time=0.846 ms

100 packets transmitted, 100 received, 0% packet loss, time 99086ms
rtt min/avg/max/mdev = 0.652/2.604/9.618/1.931 ms

hey mikrotik staff could you help me?

[gringoZ]

i found this: http://www.babinszki.com/Networking/Max-Ethernet-and-TCP-Throughput.html,
click “Skip ahead to the TCP/IP on Ethernet Performance Model” and download excel spreadsheet,
open it, change “select Line Rate:” from “Gigabit Ethernet” to “Fast Ethernet” and look at

it say that with rtt 3ms max throughput is 60mbit.
i had rtt 2.7ms, but wireless link 802.11n is halfduplex.

if this is the reason that I cant reach better throughput that 40mbit with ookla speedtest mini at local server?

how i can improve rtt at my link? how i can improve my link?

1. try mt with gigabit lan?
2. buy mt with faster cpu, is that affect al throughput?
3. something other…

maybe this model of tcpip not refer to mt wireless link? (isnt impact to nstreme?) i am wrong with it?

[normis]

first of all, you are testing with a single TCP connection. this will never show true performance of a link. you must use 10-20 connections. simulate this with a torrent download or 20 FTP connections.

Also, where is the speedtest hosted? if you test through so many hops, you are not testing the sxt link speed, you are testing your network speed. in this case, if sxt has 300mbit rate connection, it means the wireless link is ok. how about the other devices there?

[gringoZ]

@normis

first of all, you are testing with a single TCP connection. this will never show true performance of a link. you must use 10-20 connections. simulate this with a torrent download or 20 FTP connections.

ok, but why i cant reach same speed. trhoughput of 300mbit half duplex wireless link is greater than 100mbit lan or smaller? its only one wireless hop…
what is problem? latency? how i can improve it?
i think that i cant explain users that 40mbit (local!) speedtest is equal 90mbit 20 ftp connection its more complicated that build my simply network

@normis

Also, where is the speedtest hosted?

please read carefully my first post and see atached picture. i attach modified picture, you should understand it all…

@normis

if you test through so many hops, you are not testing the sxt link speed, you are testing your network speed. in this case, if sxt has 300mbit rate connection, it means the wireless link is ok. how about the other devices there?

like you see at attached picture its ONLY ONE wireless hop and 2 standard cisco switch and one linux server with ookla speedtest mini.
its works ok without wireless link on mikrotik, with wireless link on mikrotik i had more than 50% worst scores at speedtest…


if the weather doesnt worse during the week i will do test at rb911g and c3550, both with gbit lan…
if i understand links that i posted below in case of gigabit lan and rtt 3ms i will have max teoretical throughput 147Mbit, it will be very funny…

[uldis]

post your /export compact printout so we could see your configuration.
Also since you have changed a lot of configuration settings I would suggest to reset the wireless interface configuration settings and configure the things you only need. Here is the command that you can use:
/interface wireless reset-configuration wlan1

What speed you are getting when you are doing UDP test between the wireless boards?