MikroTik

One of our customers sent us this test result. We have similar results, and are happy to see that customers get the same:

Is there any tools can show the actual TCP traffic as the UDP always not so accurate...
Like when this case tested is 100++ Mbps, but if real traffic flow, i dun think it can go for that speed...

it depends on what you mean by "real traffic". it's different for everyone. that's why we use UDP test so that results are consistent and can be compared.

Please post test in TCP and settings of the wireless

It's not our test, it's from a chinese forum.

I understand. But I would like official test from Mikrotik

UDP is always the better looking test for MT to use as it always shows results far higher that what you can actually get in reality

UDP is always the better looking test for MT to use as it always shows results far higher that what you can actually get in reality

not true! it's only reasonable to test the absolute maximum possible. there needs to be some measuring point where all the tests can be compared, you can't run speedtest.net and expect it to be even remotely accurate, or even to try and compare that result to other routers. Not all web traffic is small TCP packets. There are all kinds of requirements and not everyone needs the SXT for online gaming for example.

UDP is always higher and higher is better for MT :P

How can you say UDP is the only one where all the tests can be compared? TCP isn't much different but does strain the routerboard a little more.

In our experience TCP test's are more in-line with what we can push than UDP test, We never see close to UDP speed results in a real ISP enviroment

same here

Good result. How far it was ? Signal levels ? Configuration ? 2chain or 1chain , 20 or 40Mhz width ?

Its would be first time for me to see 2chain stable link. May be we are doing something wrong, but using 1 chain we got always very nice stable links on 20MHz od 40Mbjt channel, but something is always wrong after going to 2chain setup. Its speed is everytimes very unstable.

Our units arived today, so looking forward to see results

Be nice if f SXTs came with a baseline configuration for PtP so we only had to change SSID and Channel. Right now there is so much trial and error involved

Obviously all environments are different but there must be a common starting point?

SXT supports Flashfig out of the box, so you can do one-click configuration out of the box.
http://wiki.mikrotik.com/wiki/Flashfig

Flashfig - new to me - cool

Waiting for my SXT's now - looks very interesting

Rod

SXT supports Flashfig out of the box, so you can do one-click configuration out of the box.
http://wiki.mikrotik.com/wiki/Flashfig

nice, but we are still missing default recommended values.
You should be politician

Here is my test result:

Site A
1. Notebook HP Probook Corei3
2. 4GB Memory
3. Broadcom Ethernet
4. Mikrotik RB/SXT
5. Firmware version 4.16

Site B
1. PC Dual Core
2. Memory 2GB
3. Realtek Ethernet
4. Mikrotik RB/SXT
5. Firmware version 4.16

Software used : JPerf version 2
Protocol : TCP
Distance : Indoor, 5 meters, different room

You wanna print the wireless config?

great.
can you post your configuration here.
thanks

Would love to know a config where we can cheive anything like these speeds, the most i can get through these units at the moment is 30mbps tcp one way or symmetric seems fine,

please someone post a usable config for these units

manners, push though them not between them using TCP, as they are not to powerful to generate so many TCP traffic. Try UDP.
Also test out Nv2 protocol.

UDP yes i understand i can see in the region of 90mbps to 170mbps however this really does not happen when used in real life situations, there are many out there that will agree its more tcp throughput we would like to see. i tried it with pc's at each end using mikrotik bandwidth tester for windows but the results were lousy, would be nice to see a config that works and performs

in real life situations there is no bandwidth test sitting on the SXT killing it's CPU. you can't compare that as well. the test should not be run on the SXT itself.

....i tried it with pc's at each end using mikrotik bandwidth tester for windows but the results were lousy, would be nice to see a config that works and performs

Anyway Normis, UDP test is not for messuring link capacity. Tcp test run on separated devices can be close to real traffic capacity.
I still have the question: doy yo plan some knid of RB811 or Rb811ah routers?
That would be an awesome for the new N generation.

in that case, put many powerful PC devices behind those SXTs and run multiple high speed TCP connections from PC to PC, not from the SXT itself.

I have no comment about future devices. Some you will see at the MUM, so you must come

Ok, I'll be there.

Anyway Normis, UDP test is not for messuring link capacity.

no, exactly opposite. UDP is measuring maximum link capacity

Tcp test run on separated devices can be close to real traffic capacity.

real world customer download speed is not the same as "maximum link capacity". you are talking about two different things. TCP has backoff algorithm (http://en.wikipedia.org/wiki/TCP_conges ... _algorithm) so it looks like it's running slower. Therefore you need many TCP streams to properly test the link.

Also read this http://tools.ietf.org/html/rfc675

I won't argue.
For me - as a provider - important to know whar traffic can my link handle without packet loss and lag.
And this can be messured by a tpc test. Even I test my lnik with udp and find that hurrah, its 120Mbps,
it probably will lag at 80mps traffic. Why? Because of the prototcol - of course. Since the real traffic is
mostly tcp, I have to make tests on tcp. And since we use rb-s and not pcs (that good old times) we can not messure
link on tcp with routerboards itsefls. Its clear.
Udp test is nice, and can show big numbers, but we do not manage udp networks, but tcp...
As a trainer I usualy presented the difference for my students. One thing is to show the max ,and an other to show the real.

What you say is true, except that if you have a large number of TCP connections, you will get close to what you can get with UDP. If you only have one TCP sessions, you will not get close to the UDP result. Normally, if you have many customers, you will get these results.

Also, we use UDP for all our tests to be consistent and to be able to compare performance between devices.

Just download 5 very popular torrents - it will be pure TCP test with descent amount of connections . Mind that number of connections have impact on conntrack performance also . and best thing is - it represents "real" traffic of my network with 95% precision

With proper QoS you can ensure that important connections doesn't suffer from packet drops while still utilizing max-capacity of the link.

Ok have retested the units again today, power turned down to 2dbs as this is in my house with around 5 meters between them, connecting at -55, using jperf i can get 90mbps -98mbps tcp throughput using 20 streams, one way, would love to know how the person above got the same sorta results symmetric, have to admit its very nice to see,

i have them set up as a basic bridge, bridging the ethernet and wireless at each end
using station bridge on station end
NV2
conn tracking is on

both chains are enabled

desktop dual core 2.8ghz, 3gb ram at end A
Laptop dual core 2.1ghz 3gb ram at end B

have to admit i couldnt see the same with mikrotik bandwidth tests

when i do the tests both ways it seems to average around 40mbps one way 30mbps the other

have tried with connection tracking off but doesnt seem to make all that much difference

also is there a way to switch off the led's

the mikrotik bandwidth test exe for windows uses only one TCP stream, so better use torrents like macgaiver suggested. run some 5 parallel downloads which have more than 100 seeds. maybe some ubuntu releases

Could Really do with a way to switch off the led's as im sure it wont be long before people start to notice them,

any help is greatly appreciated,

as for downloading torrents not so easy at home as the connection i have is only 20mb itself, and also as an isp in the office not possible as this could disrupt other customers, personally i cant find any faults with jperf bit of a pain to get it to work on 64bit os but other than that its worked a treat

Many thanks

Check the 4th page of user's guide.
http://routerboard.com/pricelist/downlo ... ile_id=280

楼主的设置好棒
能加你QQ指导我一下设置吗？

what about 150+mb/s tcp? I achieved with SXT in lab environment with NV2