Hi Vijay,
on 2006-05-18 20:45 Vijay Devarapalli said the following:
> Henrik Levkowetz wrote:
>> Hi Vidya,
>>
>> on 2006-05-18 01:59 Narayanan, Vidya said the
following:
>>> All,
>>> There has been some discussion lately about the
use of MPLS LSPs to
>>> carry NETLMM signaling. Traffic engineering has
been quoted as a reason
>>> to do this. I'm trying to understand the
vision behind this and what
>>> kind of traffic engineering benefits people see
this offering.
>>>
>>> I agree with Vijay's statement that TE would
lead to per-MN LSPs and
>>> that isn't exactly scalable.
>
> I should clarify here. I had said if you want to treat
> traffic to a particular mobile node in a special way,
> then you would at least one more LSP. folks were
> claiming that you have one LSP between the LMA and the
> AR and traffic to all mobile nodes currently attached
> to the AR would go through that LSP.
Right. One LSP would mean the same treatment for all
traffic.
>> This isn't immediately obvious to me -- why would
this be so?
>>
>> Having for instance 3 different traffic classes
would, in my
>> understanding, lead to 3 LSPs between local
mobility anchor (LMA)
>> and mobility access gateway (MAG), rather than
per-MN LSPs.
>
> this would work if all the sessions to all the mobile
> nodes attached to the particular MAG (a new term?)
could
> be mapped to these three LSPs you have between the LMA
> and the MAG.
Exactly.
Henrik
|