[Mobopts] Re: [Mip6] Delay Analysis for Handoffs with Mobile IPv6 Route Optimization

> We analysed these properties for the case of FMIPv6 (which should be
> transferable in major parts, e.g., fmip needs fback, you need BU to
> arrive ...) and found no significant improvement by predicitions (see
> bib-reference attached).

Thomas,

thanks for bringing this up.

While your results [1] on FMIPv6 and HMIPv6 shed light on the
performance of proactivity in combination with local mobility
optimizations and are as such undoubtedly very important, I do not
necessarily consent with your statement that those results can be
conveyed to the end-to-end mobility management that we have been
analyzing [2].

During an end-to-end Mobile IPv6 handoff, communications are delayed for
at least a global RTT between the MN and the CN (assuming that
end-to-end optimizations such as EBU and CBA are deployed).  This
minimum delay can basically be eliminated in two ways:

- Proactivity, i.e., send a BU from the old location and move to the new
location only when packets can expected to have arrived there.

- Local mobility optimizations such as FMIPv6 and HMIPv6, which either
eliminate global signaling (HMIPv6), or move it off the critical
blackout period (FMIPv6).

Your work [1] deals with the second option, our's [2] deals with the first.

To be more explicit, let's assume we use FMIPv6.  All global signaling
is then deferred until the MN has already resumed communications, from
the new link, through the AR-to-AR tunnel.  This is true for both
proactive and reactive FMIPv6 modes.  What *FMIPv6 proactivity* saves
you is just an FBU's propagation time from the new AR to the old AR.
Compare this to the global RTT that *end-to-end proacitvity* potentially
saves you.

Referring to the problems with predictive handoff procedures that you
mention:  You are correct, and section 7.1 in our analysis duely
discusses this.  As far as unreliability in predictions goes:
Unfortunately, you are right, but that's a different construction site. ;)

Regards,
- Christian

[1] http://www.informatik.haw-hamburg.de/~schmidt/papers/telesys05.pdf
[2]
http://doc.tm.uka.de/2006/vogt-2006-delay-analysis-for-reactive-and-proactive-handoffs.pdf

-- 
Christian Vogt, Institute of Telematics, University of Karlsruhe
www.tm.uka.de/~chvogt/pubkey/

Thomas C Schmidt wrote:
> Dear Christian,
> 
> thanks for bringing your technical report to our attention!
> 
> For a comment: it is our experience that the optimistic concept of zero
> handover delay due to predictive procedures does not survive reality.
> 
> This is mainly due to
> 
>   - the independence of anticipation, signaling and handoff in
>     occurrence and timescale.
> 
>   - predictions turning out as rather unreliable.
> 
> We analysed these properties for the case of FMIPv6 (which should be
> transferable in major parts, e.g., fmip needs fback, you need BU to
> arrive ...) and found no significant improvement by predicitions (see
> bib-reference attached).
> 
> Song et al. ([11] in our references) made empirical studies on
> predicitions including the campus geometry, statistical correlation
> analysis of user date ... and arrived at dissapointing 72 %.
> 
> It's a nice vision - but sometimes real-world isn't as nice ;)
> 
> Best regards,
> 
> thomas
> 
> Christian Vogt wrote:
> 
>> I would like to call your attention on work that we have done on IPv6
>> handoff procedures:
>>
>>   "A Comprehensive Delay Analysis
>>    for Reactive and Proactive Handoffs
>>    with Mobile IPv6 Route Optimization"
>>
>> http://doc.tm.uka.de/2006/vogt-2006-delay-analysis-for-reactive-and-proactive-handoffs.pdf
>>
>>
>>   Abstract:  Optimizations to reduce handoff delays
>>   inherent in Mobile IPv6 Route Optimization as well
>>   as IPv6 router discovery, address configuration,
>>   and movement detection have so far been mostly
>>   considered on an individual basis.  This document
>>   evaluates three integrated solutions for improved
>>   handoff experience in surroundings with different
>>   preconditions:  reactive handoffs with unmodified
>>   routers, reactive handoffs with router support, and
>>   movement anticipation and proactive handoff manage-
>>   ment.
>>
>> Your feedback will be welcome!
>>
>> Regards,
>> - Christian
>>
> 
> ------------------------------------------------------------------------
> 
> @article{sw-pvrah-05,
>   author = {Thomas C. Schmidt and Matthias W{\"a}hlisch},
>   title = {{P}redictive versus {R}eactive -- {A}nalysis of {H}andover {P}erformance and {I}ts {I}mplications on {IP}v6 and {M}ulticast {M}obility},
>   journal = {Telecommunication Systems},
>   volume = {30},
>   number = {1--3},
>   pages = {123--142},
>   month = {November},
>   year = {2005},
>   publisher = {Springer},
>   address = {Berlin Heidelberg},
>   url = {http://dx.doi.org/10.1007/s11235-005-4321-4},
>   abstract = {Handovers in mobile packet networks commonly produce packet loss, delay and jitter, thereby significantly degrading network performance. Mobile IPv6 handover performance is strongly topology dependent and results in inferior service quality in wide area scenarios. To approach seamless mobility in IPv6 networks predictive, reactive and proxy schemes have been proposed for improvement. In this article we analyse and compare handover performance and frequencies for the corresponding protocols, as they are an immediate measure on service quality. Using analytical methods as well as stochastic simulations, we calculate the performance decreases originating from different handover schemes, the expected number of handovers as functions of mobility and proxy ratios, as well as the mean correctness of predictions. In detail we treat the more delicate case of these rates in mobile multicast communication.
>  It is obtained that performance benefits, expected from simple analysis of predictive schemes, do not hold in practice. Reactive and predictive handovers rather admit comparable performance. Hierarchical proxy environments -- foremost in regions of high mobility -- can significantly reduce the processing of inter--network changes. Reliability of handover predictions is found on average at about 50 \%.},
>   theme = {mipv6|mmcast},
>   file = {http://www.informatik.haw-hamburg.de/~schmidt/papers/telesys05.pdf}
> }

_______________________________________________
Mobopts mailing list
Mobopts@irtf.org
https://www1.ietf.org/mailman/listinfo/mobopts