Community discussions

MUM Europe 2020
 
rpingar
Long time Member
Long time Member
Topic Author
Posts: 541
Joined: Fri May 28, 2004 2:46 pm
Location: Italy

v3rc and poor performance about nstream-dual in real field

Thu Oct 04, 2007 12:29 pm

i am testing nstream-dual on rb333 and v3 from rc4 to rc6.

It is a lkink of about 8.5km with perfect los wuth a dual pol MTI panel antenna of 35dbi gain.
We have very poor performance. Around 7mpbs fullduplex.

In labs we reached 25mpbs FD. But in real field we have the following problems:
- the rate fluctuate very erratic;
- the signal level fluctuate very erratic, from -59 up to -75;
- sometime the rb333 after an nstream-dual configuration change reboots.
- when there is no traffic the tx/rx rates is stable at 36, the ping time is fine 2ms/1ms

I tried to test separately (using nstream1) the two wlans and they works ok, with same frequencies used in nstream dual i have 15mbps on both directions.
Also i tried to test the possible isolation between the two pol af the antenna, running btest on one and pinging the oher side of the other wlan, without any apprecciable problem.

Any other experience?
regards
Ros


Any other experience
 
uldis
MikroTik Support
MikroTik Support
Posts: 3428
Joined: Mon May 31, 2004 2:55 pm

Re: v3rc and poor performance about nstream-dual in real field

Fri Oct 05, 2007 11:33 am

could you send the support output files from those units when they perform poor and send them to support@mikrotik.com
 
rpingar
Long time Member
Long time Member
Topic Author
Posts: 541
Joined: Fri May 28, 2004 2:46 pm
Location: Italy

Re: v3rc and poor performance about nstream-dual in real field

Fri Oct 05, 2007 5:45 pm

sent you the supout files.
Please check the ticketing system.

regards
Rosario
 
rpingar
Long time Member
Long time Member
Topic Author
Posts: 541
Joined: Fri May 28, 2004 2:46 pm
Location: Italy

Re: v3rc and poor performance about nstream-dual in real field

Mon Nov 26, 2007 6:01 pm

i tested the dual stream in real field with two very well separated antenna and still have the same problem.

The performance are very poor. The signals are at around -60 / -62. Un link is on 52xxmhz the other is on 57xxmhz.

When there is no traffic the ping is 1-2ms and the modulation is 48/54mhz
when the traffic start to flow the modulation goes grazy and the ping skyrocketed to 900ms. The max with udp in both direction is 5mbps.....

each link has a very impressive performance using nstream1, bonding them using eoip let it to the same nstream-sual issue.

The only thing i can imagine is that the mini-pci connector is not good to handle two cards that have to work togeter.

Ros
 
pd
newbie
Posts: 26
Joined: Thu Jan 12, 2006 10:46 pm

Re: v3rc and poor performance about nstream-dual in real field

Mon Nov 26, 2007 7:48 pm

Did you try to adjust the power of the cards? Are they at the default setting?
I had an issue of signal fluctuating MUCH within milliseconds when I tried to lower
the transmit power of an XR5
With default power it was OK
 
Skaught
Member Candidate
Member Candidate
Posts: 146
Joined: Mon Jun 19, 2006 9:31 pm

Re: v3rc and poor performance about nstream-dual in real field

Mon Nov 26, 2007 8:41 pm

We too have had issues with real world deployment of Rc10 on RB333.

Our link is almost identical to yours in terms of signal and distance.

We first took out the old link, a Star-os WAR. We "upgraded" to a single card RB333 solution. Speeds and latency instantly got WAY worse. Dial-up speeds and latencies well over 200ms. Even with one end on Star-os and one end on MT, the link was garbage. We put it back on Star-os, link was great. Same channels, same dishes, same feeds, and same model of wireless cards (UBNT SR5).

We started banging into the 28mbit limit on the Star-Os WAR boards and wanted the ~40mbit we saw on the RB333 in our shop so we put the RBs back in again. Problems instantly returned. We fought with it for hours until in desperation we turned on nstreme with polling. Link was great!

Now we are running out of bandwidth on the single card link and want to move to bonded interfaces and ditch nstreme/polling. Only thing is I have ZERO faith that the product will work. Works great on the bench, but as we saw before that means nothing.

I am not really interested in nstreme2 as our traffic is HIGHLY asymetrical which means running full duplex will not buy us much. We already push over 30mbit in downstream. Full duplex would only boost that about 25%.

My many years and thousands of deployed links (I have been doing this longer than anyone else in this part of the world) tell me there is a bug someplace. When I pull one set of gear and drop in a new one and the difference is this clear, it is a no brainer. Unfortunately MT has been less then helpfull on this issue for me. I doubt there are many users pushing the kind of data we do at many thousands of PPS.

Who is online

Users browsing this forum: Sapo and 125 guests