The MPLS-OPS Archive

["Alexander Marhold" <alexander@marhold.at>]

>Variable length packets versus fixed length cells or TDM timeslots is what
>leads to the jitter problem on packet based circuit transport.

I slightly disagree, as without QOS (queueing,PHB) mechanism
it gives the same JITTER if you on a packet based network for two
consecutive packets have
a  100Byte and then a 1500 byte packet in the queue before you
or on a cell based system you have for the same 2 packets the same data now
as
 3 cells and then 45 cells in front of you.

so ATM by itself does reduce only the serialization delay of a packet going
through (as the delay is one time the full packetsize serialization delay
and for all other ATM-hops only the serialization delay of one cell)
but it does not reduce JITTER, when other variable amount of cells are in
the queue in front of you

Following the discussion I also have the impression that the old myth of
"ATM is QOS" and "IP is best effort without QOS" is somewhat behind the
arguments.

IMHO that is not true as "ATM with QOS" people think about CES ( circuit
emulation services) or similar and for IP they think of Internet.
In reality for the same given traffic streams which are today mostly IP, the
same QOS can be achieved in routed L3-networks( with or without MPLS) and in
core-ATM networks (the edge and customer is frame based, as ATM to the
desktop is not the case)

On the other hand 2 remarks on MPLS:

1.) nobody is forced to use it neither in the SP nor in the enterprise arena

2.) MPLS and MPLS/VPN has also some issues you need to consider when
implementing:
    a) a labelled packet looses its identity, which means without
classification you cannot distinguish a voicepacket from a datapacket
anymore as the L3-header is unacessible within the MPLS-core
    b) In peer-to-peer networks like MPLS/VPN, BGP and redistribution leads
to higher convergence times than in L2 overlay networks using FR or ATM
tunnels.

But have you ever set up a network with 2300 endpoints, 200 distribution
points and any-to any connectivity for the 200 distribution points (it means
200*199/2 circuits in FR and/or ATM)

with best regards

Alexander Marhold
Senior Consultant
PRO IN

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