Question Regarding Beacon and DTIM Intervals
Last Post: March 17, 2006:
-
Joel,
In a post in a previous thread http://ww.cwnp.com/phpBB2/viewtopic.php?t=468&sid=41f4f29c0a55323d73b36ea7778e7a73
you said:
When I analyzed the WLAN, I noticed that beacons were being sent every 10ms instead of the normal 100ms...
Is there a best practice (with explanation for why) that recommends what value to set the Beacon Interval and DTIM Interval? Judging from some posts on this site and from some device documentation, 100ms appears to be a common value for the Beacon Interval. I have seen values of 1 and 3 for the DTIM interval. What would motivate an administrator to adjust these values?
Thanks,
moe -
Hey Moe.
While this might not be a complete answer, I have learned that adjusting the DTIM interval can be important if you have VoWLAN apps in use. That kind of data can't wait around in a buffer too long. Drawn out DTIM intervals and a good amount of activity in the BSS could cause problems for the VoWLAN user(s). One solution would be to decrease the DTIM interval. I suspect streaming audio &/or video would fall in the same boat.
I too have seen default values at both 1 and 3.
-Tim -
wk4u,
Thanks for the response. It is still a bit unclear to me.
Let me spill my brain here..a Bounty quickerpickerupper should suffice..
TIMs are used by devices for power save mode allowing them to know when to wake up and check for queued unicast traffic;
DTIMs are used to signal queued broadcast/multicast traffic;
I can see where the TIM intervals would be important for unicast streams and VoWLAN but I don't see the connection to the DTIMs unless the traffic is multicast. I am not very familiar with VOIP but I assume it is unicast.
Regards,
moe -
Hey Moe.
I misspoke when I mentioned VoWLAN. What I should have said was streaming audio &/or video. You're correct that VoIP is unicast traffic. Heck, I suppose it could also be hosted on 802.11 via UDP? Up to the vendor I reckin.
I think you're correct about sleeping clients watching for TIM frames. They all know to listen about 10 times a second. IF the client's AID is in a TIM frame, then he knows to stay awake and request that unicast data from the AP.
Your question concerns the DTIM though. As I understand it, if I spread the DTIM count out (for example, 20), any buffered multicast data from the DS might not be delivered to the clients associated to the BSS in a timely manner. Especially if there is a good amount of activity within the BSS. If any of the clients are running an application that needs the multicast data timely, a large DTIM value would hurt their performance. I suspect the flip side to a large DTIM value would be higher unicast throughput within the BSS? Maybe we can get somebody else with more experience to chime in and enlighten us more?
-Tim -
Hey Moe.
I'm surprised nobody added anything to our little chat yet. Anyway, I decided to try this stuff out here at home. I guess it was the dance music from the Internet that motivated me?
Turns out even drastically changing the values of the beacon interval or DTIM interval have very little effect on throughput. I noticed less than 50 Kb in throughput change. Now if you really want to make a difference in throughput, force your AP to use short preambles (if all your clients will go short) and watch the througput go up over 500 Kb!
I suspect AP manufacturers include the beacon & DTIM interval settings - and others to thoroughly confuse end users and force them to hire CWNP certified professionals at outrageous rates to fix their problems.
-Tim -
You are tempting me to read a thread I have intentionally avoided while I try to dig deeper myself. I am talking about the trace question that wirelesswizardCWSP posted regarding use of the long preamble when the BSS was capable of using a short preamble.
But back to the DTIM and beacon intervals.. I have not had any time this week to research this..does anyone out there know a WiFi software designer who can give us insight into the values that have been commonly selected for default Beacon and DTIM values? -
This is a topic I also allude to, but don't firmly justify - I have concluded that the environment is the bottom line issue - sort'a like - frag values may help a particular situation, but you can't post a firm "use this value". I know one client with Cisco voice over wireless, and the unit would not function properly until DTIMs were set to a value of "1". Was it because such a setting forced the clients to be more responsive? Or, was it enabling the AP to be more responsive because there was no need to buffer data for a length of time? It's hard to say. The same goes for your beacon intervals - the timing is important to the network - average values are as you mentioned - but, specific environments may need /or/ benefit from tweaking.
- 1