The MPLS WG Archive

[Dieter Beller <D.Beller@alcatel.de>]

Vijay, Ping and all,

see my comment below.

"Vijayanand C - CTD, Chennai." wrote:

> >Now, back to RSVP, when it is used to provide QoS signaling for
> >inter-active real-time sessions (such as voice), it does not work well.
> >This is mainly because the duration for real-time sessions is normally
> >short. An average phone call is 3-4 minutes. When you plug in the
> >network topology of an existing SS7 network, and the desired goal of the
> >3G networks, the number of RSVP session changes (on/off traffic sources)
> >is in the range of a few thousands sessions *per second*. When using
> >RSVP, routers will be under a lot of stress. In this case, RSVP is not
> >sufficient, refresh reduction has no use, and the soft-state approach
> >seems to be a joke. I think this is the problem that NSIS WG is trying
> >to handle.
>
> This is precisely the kind of scalability/performance issue that I ( and
> probably nagabhushana) were pointing to. Soft state is not scalable always.
> This problem is likely to be acute when  GMPLS networks are fully deployed.
>
> Vijay

I disagree with the statement being made for GMPLS above.
GMPLS is still a brand new technology in transmission networks
and is intended to make them dynamic. But the dynamics you
are describing above will not come over night. This will be the
result of a rather lengthy evolution from a quasi-static situation
which we have today to a fully dynamic network in a couple of
years - and I'm not sure whether we will arrive at a situation
 where thousandsof large pipe connections will appear and dis-
appear within a second. This will definitely not happen in the
backbone because traffic is expected to be highly aggregated
there.

Actually, I'm in favour of the proposal how to move forward with
RSVP-TE (standars track) and CR-LDP (informational).

Cheers,
Dieter
begin:vcard 
n:Beller;Dieter
tel;fax:+49 711 821 - 42303
tel;work:+49 711 821 - 43125
x-mozilla-html:FALSE
org:Alcatel - Optics Group - Terrestrial Networks Division;US/EA3
version:2.1
email;internet:D.Beller@alcatel.de
adr;quoted-printable:;;Alcatel SEL AG,=0D=0ALorenzstrasse 10;Stuttgart;;70435;Germany
fn:Beller, Dieter
end:vcard