I have a RB433 with ROS v5.11 installed. The internet comes on a wireless 5ghz PtP link and there are no problems with it, the throughput and latency are stable at all times.
There are clients connected on an omnidirectional antenna, around 15 clients, their signals are -60dBm ~ -75dBm, ccq is between 70 and 100. The wireless card is Atheros AR5213.
The problem is that the latency goes to 1-2sec and at that time the throughput can not go over 1Mbit/s. If I disconnect all clients so they can connect again all goes no normal - latency is 2-3ms, the throughput can go to 10-12Mbit/s. But after some time that varies from 2min to 30min all goes bad again, reconnecting the clients immediately solves all problems for another short amount of time.
I added a rule to limit connections to 35 on every user. It improves things a little, but still the problem exists. I have 300-400 Tx and the same number Rx packets.
i have same problum
i m useing too rb433 with mini pci R52h
same situation when users more then 20 online
my through put rate stable but Rx rate decreesing at 1mb when latency going high like 200ms 400ms
plz help
20 clients? I now see it.
I think, if you would like to work this normally, you have to reduce your clients maximum 10/AP in 2,4Ghz b or g mode without TDMA.
You try to enable hardware protection rts/cts on AP and clients, set up the rts threshold, and the frame fragmentation (Hidden node problem and collission).
It is not enough just on the AP!
Watch the ack times in registration table and if it is too large (greater than 100) or not a stable value, maybe the client is bad.
I have tried using rts/cts but it changes nothing. I did not however played with the threshold value. Can you advice on some value ? The best throughput is when I use g-only standard with full auto on tx-rate. I would like to use nv2 as i know it will improve things a lot, but all clients are tp-link.
The thing that helped me most so far is limiting to 3 udp connections (!53) and 10 tcp connections on 1024-65535 ports.
Still the latency spikes to 1sec for about half an hour and then goes back normal.
Also I wonder if frame lifetime can help with things ?
Try to set the rts threshold to 512 in the AP, and in TP-Link clients. The frame life time…I don’t know exactly, I think it’s a store time, so how many times will be stored a frame if the router can’t send it instantly, before drop them.
I don’t know the value is milli-, mikro- or nanosec for frame lifetime. This is not sure to help you, but try it.
But my opinion, without TDMA 10 active clients are the maximum (outdoor) in one 2,4Ghz AP.
Just a slightly different perspective, I’ve got 31 clients on a single R52H, straight 802.11G. This AP is going to be upgraded soon, but the only problem I have with it now is if someone is using torrents, which caused the same problem back when there were only 5 people on it.
Sounds like you may have a problem with your card or a bad client. Have you run torch to see what is going across the interface during these problems?
Bittorrent will send thousands of UDP packets per second. The data rate will only be ~100k, but it will thrash an access point.
CyberTod, would you mind posting your connection limit rules? I have tried several different methods and had no luck.
Almost all of my clients use torrents I have an AP with ~30 clients too, the situation is almost the same. So I guess the torrent users do more damage than connected clients.
I saw this behaviour a couple of times on a 2,4AP (~30 CPEs), suddenly ping times rise from 1 .. 5ms to > 500ms and the only way to make it back was those turn around already mentioned: make all clients reconnect back.
I tried some setups to keep it stable and at least at a 20MHz BW channel on R52nM cards I settle down using only G mode, legacy speed selection, configured to only G speeds, distance = dynamic and adaptive noise immunity = off. No hw protection.
HW retries = between 4 and 7. Higher than that = higher ping times.
I use to run adaptive noise immunity = on and distance = 2Km 'til ROS 5.0X but after - I don’t remember - 5.5 under such setup I started to face this ping time jumps.
To be honest no conclusion except I keep adaptive noise off, hw retries < 7 and data rate = legacy on 20MHz channel. On 10MHz channels distance = 4Km, adaptive noise immunity = on, only G mode and configured data rates (legacy).
This AP uses 3 freq on 2.4 (is a 433UAH with 3 x R52nM) and handle ~30 CPEs each frequency (access list have 92 UBNT CPEs).
ur problem is HIDDEN NODE n u cant do anything about it untill u shift to tdma. some of your clients cant see each other so if they start transmission together collisions will occur at ur AP (ur AP can listen to one client at a time)
here are some suggestions
-ur AP can see all the clients so turn on the “CTS to SELF” on AP
-all of ur clients cant see each other so turn on rts/cts on ur cpes ( i m also using tp-link products which are realy weak but cheap)
-while installing ur client end devices try to give them as much height as possible so they can see each other
Finally dont put emphasis on a large ptmp networks (if u want weak and cheap cpe to work). instead of one AP add many APs
ur life will b easy
So what is better in your opinion - cts to self or rts/cts. Yes the clients do not see each other very well so it very well could be hidden node problem. I have tried setting rts/cts on ap and 512 threshold. On the tp-link clients there is no way to turn it on/off but there is a threshold value so i guess it is always turned on. The default value is 2346. I have changed it to 512 too, but not on all devices. Is it a problem if on some of them it stays 2346 and on others 512 ?
Also i have enabled wmm just to see if it makes any difference. I dont see improvement, but dont see it going worse either.
We have for redundancy dubble setup, 2 x 2.4 for clients, 2x5 Ghz for connection with central location, normally on separated 433AH.
For testing I try to transfer some clients on “backup” AP, but if only 2 clients are on AP, problem persist
I have news about high latency…but I don’t know what is reason for this behavior.
I power up Snooper on another card, and see that usage of Freq. are high on this time high ping, but is not visible that is any client using it, because is on same MAC as AP card is ?? Picture in attach !
What might this be ?
I now see it.
