Re: On Link Triggers (implementation)

On Thu, Sep 26, 2002 at 09:01:02PM -0700, Behcet Sarikaya wrote:
> Hi Jean,
>   Thanks for posting this information. 
> My question is, you state that it takes about 1sec to
> complete link layer handover which I found very long
> and somewhat discouraging if you consider real-time
> communication such as voice-over-IP. 

	I wasn't talking of the handover time, I was talking of the
time to get a trigger. And it's not always one second, that's an
oversimplification.
	In details :
	o IrDA : connection blocked trigger will be one second.
	o 802.11 managed : The trigger is NEWAP, which is the end of
the L2 handoff. The L2 handover may take up to 1s (probe_time *
num_channels). You probably should do your V-Handoff only when you are
sure that the L2 handoff did fail.
	o 802.11 ad-hoc : first trigger will probably be in the order
of 100ms (using TXDROP, (contention_time + transmission_time) *
retry_max - assuming there is active traffic to that node). To avoid
false positive due to long fade, you may want to wait for a few events
before making the final decision.
	If you are actively monitoring the SNR on 802.11, you may be
able to predict the trigger (but not always).

>   Do you think that during this period no IP layer
> activity can take place?

	Nope. The trigger above come at a point where you have already
lost connectivity (for the time I describe above).

>   If yes, then I think what Mobile IP WG is trying to
> do to have fast Mobile IP handovers is simply
> impossible, e.g. the mobile trying to inform the old
> access router of the new access router, etc.

	You can do better. For example, on 802.11 you can monitor the
SNR to predict the trigger in advance. But, this require to be very
intimate with the characteristics of your radio receiver, and there is
a nice tradeof between advance time and false positives, so this would
have to be set in function of the required QoS. And also you have a
nice interaction between L2 and L3 handoff.
	If you have regular Internet traffic (HTML, FTP, media
streaming and co), the 1s trigger above will be plenty enough. Only in
the case of VoIP you need to do better.

> --behcet

	Have fun...

	Jean
_______________________________________________
pilc mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pilc
http://www.ietf.org/html.charters/pilc-charter.html
http://pilc.grc.nasa.gov/