Change DTIM settings?

[RicN]

I have some questions on the DTIM period setting, available on the Procurve Access Points and WES module.

1. Does the wireless stations have to be configured/have support for the very short sleeping periods you get from DTIM?

2. For a laptop going on batteries, is there any noticable gain in battery lifetime by using DTIM at all?

3. Do you ever change the DTIM period to longer/shorter? I am thinking if question 2 is yes, then would it be of value to raise to DTIM period to 3 or 4 perhaps?

If you have any answer or good guess on any of the question I would be very grateful!

[RicN]

I think I misposted this, I guess it should be in the wireless section. Sorry.

[Manfred Arndt]

1. Does the wireless stations have to be configured/have support for the very short sleeping periods you get from DTIM?

Yes, the appropriate power saving mode has to be configured for the WLAN interface on wireless stations to gain any power saving benefits. Stations not confured will still work just fine...just won't benefit.

2. For a laptop going on batteries, is there any noticable gain in battery lifetime by using DTIM at all?

Yes, there is a benefit as this causes the radio to be turned off when not sending traffic...this is especially useful for wireless IP phones that are constantly in standby waiting for a call.

3. Do you ever change the DTIM period to longer/shorter? I am thinking if question 2 is yes, then would it be of value to raise to DTIM period to 3 or 4 perhaps?

Using too large of a DTIM is not recommended. This is because the AP/RP must buffer broadcast, multicast, and packets all sleeping clients until the next DTIM interval. Using too large a DTIM can cause buffer overruns in the AP/RP and introduce latency.

A DTIM of 3 is a good default.

[RicN]

Thanks for your answers!

>Yes, the appropriate power saving mode has
>to be configured for the WLAN interface on
>wireless stations to gain any power saving
>benefits.

Is it common for Wireless NICs to have this feature and if they do, is it generaly turned on or off?

>Yes, there is a benefit as this causes the
>radio to be turned off when not sending
>traffic...

If I understand it correctly the sleeping station still has to wake up every beacon interval (normaly 10 times a second) to recieve frames and the DTIM feature makes the AP/RP hold some frames waiting to allow longer "between-beacon-sleep". :) What I am wondering is - since this is quite short intervals - is there any approximate number of battery lifetime gain using DTIM? I know that is impossible to say anything specific, but could it be that you save around 5% lifetime or 50% or.. (For a station that during that time does not send/recieve much other traffic).

>This is because the AP/RP must buffer
>broadcast, multicast, and packets all
>sleeping clients until the next DTIM
>interval.

Do I understand it correct that when using DTIM the AP/RP only caches broadcast/multicast and never any unicast traffic?

[Manfred Arndt]

>> Is it common for Wireless NICs to have this feature and
>> if they do, is it generaly turned on or off?

Yes, DTIM support is fairly common on WLAN drivers when power management is enabled.

>> If I understand it correctly the sleeping station still has to wake
>> up every beacon interval (normaly 10 times a second).

Nope...with DTIM enabled, a sleeping station only wakes up ever Nth beacon. So hence the power savings.

For example, a DTIM of 3 on an idle WLAN interface only requires 1/3 the power for the radio.

The overall system savings is based on the ratio of power consumed by the radio to the total system. For a handheld devices this can be significant. For an idle laptop, it's likely in the 10-15% range. I turn my radio off when I don't need a connection and notice a significant batter life improvement.

Here's an interesting read on the power consumed by the various components on your PC:
http://www.codinghorror.com/blog/archives/001099.html

>> Do I understand it correct that when using DTIM the AP/RP only caches
>> broadcast/multicast and never any unicast traffic?

When using DTIM the AP/RP buffers all broadcast/multicasts (if there is any sleeping station) and all unicast traffic addressed to sleeping stations.

[RicN]

>Yes, DTIM support is fairly common on WLAN
>drivers when power management is enabled.

With "power management", is that that the laptop is using some kind of power management scheme at all, or do I have to specificly enable DTIM on the WLAN card?

That is, if the WLAN card/driver has support for DTIM at all, will it automaticly use it if I enable power management on the PC?


>Nope...with DTIM enabled, a sleeping
>station only wakes up ever Nth beacon. So
>hence the power savings.

Ah, interesting! Does the AP/RP communicate the DTIM interval in ordinary beacons? (And does the beacon itself specify the beacon interval?)

>When using DTIM the AP/RP buffers all
>broadcast/multicasts (if there is any
>sleeping station) and all unicast traffic
>addressed to sleeping stations.

So the AP/RP knows if any associated station is sleeping at all and holds all broadcast traffic if so, but also caches all unicast traffic for each of sleeping stations? Very interesting!

Does the AP/RP tells these stations that they have frames to collect?

Is the DTIM a seperate control frame by itself, or is it always a part of a normal beacon?