Sorry for repost the problem, but if I can get a solution I´ll have to abandon the Mikrotik system…
I’m with problems in just one tower. In some random moments the speed of the tower is terrible… In good moments I’ve a ping average at about 7ms or in the maximum 15ms but when I’ve speed problemsthe ping average up to 400ms or until 600ms and anyone can have a good speed…
In terribles moments if I drop the P2P traffic the speed improves. It doesn’t arrive to the normal values but it arrives very close of this.
A strange think that I observed in this tower is that the signal strengh of some clients oscilates a lot. For example: a client has -60 of signal strengh when suddenly the signal goes to -78 and some seconds after it comes back to -60 again… I saw this occuring with 8 or 10 clients.
I’ve in the maximum 45 clients registred simultaneously on this AP.
See below the configuration of hardware and software tha I’m using:
I´m with the hardware configuration below:
-Mikrotik 2.9.11
-PCMCIA Senao SL-2511CD PLUS ext2 card (Prism) doing AP in 2.4Ghz;
-Dlink DWL-AG530 PCI card (Atheros AR5213) doing PTP with my office in 5.8Ghz;
-Intel Pentium IV 1.8Mhz processor
-256Mb RAM
-Flash IDE 256Mb
The software configuration:
Two interfaces are in a Bridge;
On AP card:
is configurated as AP Bridge;
I’m not using Default Authentication;
I’m not using Default Forward;
I’m using default data rates (but I tried all possible configurations with this;
I’m using default tx-power-mode (but i tried a lot of configurations with tx-power-mode and tx-power);
I’m using omni directional antenna of 8dBi (but I tried an antenna of 10dBi and of 15dBi);
On PTP card:
It’s configurated as wds station;
2.4Ghz
I’m using hyperlink 22dBi 5.8Ghz directional antenna;
I’m doing the bandwidth control on the Mikrotik with simple queues
I’ve rules to protect the client of the most comon virus;
What I did:
I did the tests with ALL CHANNELS;
I did the tests without bandwidth control;
I did the tests without firewall rules;
I did the tests without access-list control;
I change ALL the equipments… when I said ALL I went to say ALL really… I change everthig… the computer, the wireless cards, the cables, the pigtails, the antennas, the nobreak, the power cable, the power font… REALLY EVERTHING!!! I changed until the electrical input!
I’m crazy with this problems… if I can’t resolve the problem until next weekend I’ll change the Mikrotik system for another one, that I don’t know which yet…
Please someone… help me to solve this crazy problem!!!
try it with 2.8.28
we had the same problems here after upgrading to 2.9.x on serveral boxes
pingtimes even climbed up to >3000ms
this problem shows up only as long as clients are trying to connect - as long as they do not transfer any traffic they get disconnectetd after a few seconds - and this goes on until there is data transfered (until this point the signal is showing -95db)
and i think as faster the cpu on that box is the lower the maximum ping responces are - on an rb230 with 3x cm9 and about 100 clients we had up to 4000ms.
so we changed back to 2.8.x and everything works fine again…
I would suggest sending your supout file to support@mikrotik.com so that they can check it over. If you think it is really the wireless support, then you can try to enable the wireless-test package under the /system/packages menu and reboot. Also, it should be easy to downgrade to v2.8.x. But first I would send a supout file to support@mikrotik.com .
I agree with you about noise level but I tried all channels and the problems persists
How can I to measure the noise level? Or how can I prevent the problems with noise on this tower?
I need to solve this urgently because I can support the phone calls of my clients tell me about this problem that I know it exists but I can’t resolve.
Why the problem is brightened up when I drop the P2P connections?
How can I downgrade to 2.8.x to test if the problems persists? Can I do this without have fisical access to the Mikrotik? I just have access to the tower throug wireless interface.
Yes, you can downgrade remotely but as with upgrade there is some risk:
Upload the old 2.8.28 packages that you need by FTP and then:
\system package downgrade
and reboot. I have found that sometimes, after upgrade, the router must be power cycled to boot up properly so this is the risk you take doing this remotely.
Sorry but I didn’t read you main post carefully first time!
45 clients on Prism card is way to many!!!
They are making a lot p/s probably and ACK si probably high as hell.
That card can not serve properly so many clients. Been there, done that!
How many p/s u see on the AP interface when you expirience low ping etc?
Also when u said u are not using Default forwarding, did you disabled it in access-list for each particular mac address?? If not users can still exchange data and u can’t see that traffic!
I do not agree that downgrade will help!
You can try that, but I’m 100% sure that problem will still be there.
Puting some other ap or OS will not help at all.
P2P blocking is not helping a lot because card is still processing packets (requests) but the firewall is one that is droping them.
So card is actually acccepting all junk traffic!
Somebody may say “hell, why do I need p2p bloking then”
Well, so u stop p2p traffic from going further and spending your precious bandwidth.
Advice, change prism for cm9 or SR2 as soon as possible, but don’t expect to much. The best thing is to put three cm9’s and 3 sector antenas (different channel) and you back with quality!
Hmm…I believe there is some “relief” for your wireless card by implementing p2p control in the firewall. TCP works in such a way that if the path is slow it will not just keep throwing packets down it. It will wait for acknowledgement and then send the next packets. So, for example, if p2p is throttled and client is uploading p2p traffic via wireless we would see a lower p/s count through the wireless interface as well as on the WAN side. Packets that are dropped in firewall of course have to go through the wirless interface first but a dropped packet usually leads to little more following it.
Than I do some tests following your message. You can see the results below:
Number of registrations: 37
Maximum Number os packets per second: 240 ( but it was a peak, the average it was about 160pps
See the result of tests:
Without any control of P2P traffic:
Just with QOS, reducing the priority of P2P traffic do PRIORITY=8 and the remain traffic to PRIORITY=1:
Now I’m doing the QOS as above and limiting the download of P2P to 256kpbs for all clients:
How you can see in the images above the pps is not very high and the results of the ping is better if I do the P2P traffic control, but it isn’t good for me because I’ve a lot of residential customers on this tower that bought a 400kbps link and they want be able to do P2P downloads at this speed.
How I said on the first message I’m doing the bandwidht control for each IP on my network using simple queues. I’ve customers with 200kbps and 400kbps on this tower. All customers have 64kbps of upload. All customers have fixed IPs.
And about of noise??? How can I see the noise that are on the area where my tower is located?
Wow, only I can say that your AP is working great, there is no noise issue just to much traffic for 802.11b in ap mode. Unfortunatelly only solution is to use more ap’s. Trust me, your AP is working great!
It differs from situation to situation, but I found working great with not more then 150 p/s.
I don’t want to dissapoint you but that’s just 802.11b, that’s how it works
Maybe you should consider 802.11g which can hadle a lot more connections.
If I change the omni directional antenna for 4 90º or 3 120º antennas how is the ideal configuration? Use the same ssid in all aps or use diferents ssids in each ap.
I block all customers to get access to the AP using the Access list. I just allow one machine to associate to the AP and the problem persists just with a machine registrared.
Why does it occour??? The pps on this situation is about 10pps.
