The MPLS-OPS Archive

[Geoff Bennett <geoff.bennett@marconi.com>]

>1.
>When label setup method of MPLS switch is traffic-driven, how many
>packets go through the conventional IP layer before that flow can be
>sent via layer-2 switch?


As I understand it, there are currently no active developments towards
traffic driven LSP setup.  The excuse I've usually heard is that this
mechanism will create an unmanageable number of "micro-flows" in the LSRs.
This was the excuse to say the ATM MPOA was not a "carrier scale" technique
three years ago.

I'm not totally convinced that this is a valid excuse because it would be
fairly easy to define a "piggybacking" function to map multiple flows onto
the traffic driven LSP using policy based techniques.  But given the amount
of work needed to get MPLS up to speed from a functionality point of view,
this is probably not a priority right now.

If such an implementation were to be built, there is no sure fired way to
know how many packets would take the "slow path".  You might be able to get
some anecdotal experience from MPOA networks.  But in these cases, the
trigger to create a short cut could be based on things like numbers of
packets to the same IP network, rate of flow of these packets, or on
specific types of packets being sent (eg. if you see a Port 21 packet, this
looks like an FTP connection setup and so if might be assumed that a number
of follow-on packets will be flowing between the source and destination).

>2.
>After a flow flow can be sent via layer-2 switch, the accumulated
>packets in IP router buffer should send via layer-2 or still send by IP
>router?

This is difficult to answer given that there aren't any implementations,
but there are two ATM documents that might give you a better idea of how
this would be done.  Both of these are available from the ATM Forum web
site (http://www.atmforum.com).

In ATM LAN Emulation, there is a flush mechanism that administers the
change from the Broadcast Unknown Server (BUS) forwarding path onto the
Data Direct VCC path.  This is to ensure that packets do not arrive out of
sequence.

In MPOA there is no flush mechanisms as such, and the handover from the
"slow" path to the "fast" path is described.

If IP packets are already queued for transmission on the slow path of an
MPLS implementation, there may be an issue de-queueing them as router
queuing implementations are not very sophisticated.

I assume your concern is in regard to out of sequence packets?  If this is
the case please use the MPOA standard as a guide as this should be very
similar to traffic-based LSP setup, and the industry has several years of
deployment experience with this technique.


Cheers,
Geoff
================================================================
Geoff Bennett                             Tel: (33) 497 21 43 62
Director, Office of the CTO               Fax: (33) 497 21 43 50
Marconi
Gaia - Bat. E                           email: geoff.bennett@marconi.com
BP 123
06903 SOPHIA ANTIPOLIS
FRANCE
================================================================

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml