Ping delay problems!

Hi folks,

We have a computer running RouterOS with three D-Link wi-fi PCI interfaces and one ethernet interface. The equipment is acting as a bridge with three access points. The APs are using channels 1, 6 and 11 respectively and have 3, 8 and 33 registered clients.

On the interface which have 33 clients, sometimes the ping responses are with about 1000 or 2500 milliseconds! We discovered that when this problem happens, it affects others clients connected on the same wireless interface. This problem doesn’t occur all the time and this is what is making all of us crazy.

After we made lots of tests, we discovered that just when some of our clients are using Internet the problem occurs. We identified one client who could be causing the problem, so we powered off the client’s radio and the problem stopped just for some time.

We think that maybe some other clients with some kind of problem on the antenna, radio or pigtail can also be generating packages with problems that could be affecting the quality of signal to the others users.

My question is: Is there any tools that could help us to discover which clients can be causing this problem? What kind of tests could we make to discover how to solve this problem?

Our RouterOS version is 2.9.6.
The machine is have an Intel Processor with 3Ghz and 512Mb of memory.
The wi-fi interfaces have Atheros AR5213 chipsets.

Thanks for any help.

Gabriel Lauter
Brazil

Any particular reason for using v. 2.9.6? - My suggestion is to upgrade and that should help.


Upgrade and then we can help.

Regards

We are going to upgrade to version 2.9.46 tomorrow. After that we will check if the problem keep occurring and post here the results of our tests.

Tnks,

Gabriel Lauter

looking forward to your results with 2.9.46

With version 2.9.46 the situation is a little bit better. The router has not been affected by clients that have a poor signal the same way it was before the upgrade but unfortunately the problems keeps occurring in some situations. Is there a way to identify clients affecting other users?

Looks like the clients for whom the Mikrotik is not using 11MB as TX rate are those with some problem on the signal. So I ask… is there a configuration that could reduce the impact that this clients are causing to the other users from the same AP?

Another question… could it be a problem related with power supply? The machine is a normal computer with a 400wats power supply. It has three d-link wi-fi PCI cards. Does it have enough power supply for that?

Thanks,

Gabriel Lauter

try setting this for wireless: hw-retries=4

we already tried hardware-retries=4 but the result was not enought.

recently someone told me that we should use fixed rates on the client and Mikrotik sides. we tried with 1mbps and 5.5mbps for basic and supported rates but the problem keep without solution.

and now… guess what… we disabled one of the computer’s interfaces and replaced with a senao radio… and… the result was great! looks that the old senao is working 10 times better then our box solution with Mikrotik.

well, I don’t wanna blame Mikrotik, in fact I’m trying to change all of our access point to Mikrotik and use The Dude for management and monitoring, but the fact is that the senao gave us a better result.

we think that maybe the problem could be in ours d-link pci cards, they have 100milliwats and the senao have 200milliwats. maybe we’ll try to use a mini/pci adapter and senao or atheros minipci cards with 200milliwats on the same box.

does someone have an opinion about this situation?
thanks,

Gabriel Lauter

hey,

we replaced our box (and our senao radio) with a routerboard solution using 200miliwats cards and now everything is working fine!!! seems that the problem was on our d-link cards that doesn’t have enought tx-power.

thanks for all,

gabriel

weak signals tend to cause the dynamic ack timeout values to creep up over time, the senao radio will help, but not if the clients have weak transmit signals back to you. if it happens again, try seting your ack timeout to a fixed value (to start, use the dynamic value of the farthest away cpe that is not having performance issues.)