RE: IETF54- Informal discussion on BC Model for DS-TE

["Geib, Ruediger" <Ruediger.Geib@t-systems.com>]

Francois,

my 2 cents in line.

Regards, Rüdiger

FLF> Neil, Ruediger,
FLF> 
FLF> Is it acceptable to you if the model:
FLF>    - can be used without preemption
FLF>    - works "reasonably" well without preemption (but 
FLF>      not perfectly ie does not achieve all the other
FLF>      goals simultaneously)
FLF>    - may involve preemption to achieve all other 
FLF>      goals?

Effective use of bandwidth, bandwidth isolation and 
limited IGP changes are the goals I look at as most 
important.

Neil> .......but I have to point out that my
Neil> observation on pre-emption was quite general 
Neil> (so not just this DS/TE case, and it can also 
Neil> apply to other network technologies like ATM...

That would hold for me too.

FLF> Let me reclarify that
FLF>   - the Russian Dolls model does not "mandate" 
FLF>     preemption. It can certainly be used without
FLF>     preemption.
FLF>   - when used without preemption, it works 
FLF>     reasonably well , but not perfectly. ie it
FLF>     achieves bandwidth efficiency and avoids QoS
FLF>     degradation. but it cannot provide complete 
FLF>     isolation.

The lack of isolation means that you'd have to 
overprovison to make sure that a reasonable amount 
of reservations of all service types is possible. 
I'm not sure whether RD without preemption provides
sufficient added value to justify the added 
operational complexity.