The MPLS-OPS Archive

[Mathew Lodge <mathew@cplane.com>]

At 01:04 PM 7/11/2002 -0400, RickGall1@aol.com wrote:
>In MPLS Traffic engineering is two separate processes, putting traffic where
>you want it (RSVP-TE and CD-LDP, ICMP, IGP).  You must be able to put the
>traffic where you want it in-order to design traffic loads.

I think there are three processes (see below).

>The second part is doing the math. The math (the design process) is very
>involved, it takes a full understanding of voice and data traffic, trends
>statistics, and traffic modeling.  There are offline tools to help with the
>offline design process.

The problem I have with this is that it mixes up the concepts of demand / 
capacity planning with network engineering. I think this is a bad thing 
because the two problems are very different and therefore are best solved 
in different ways. Demand / capacity planning is the process of figuring 
out what the traffic forecast will be for the network in terms of how much 
traffic, and sources and sinks of that traffic. So, this is when you decide 
that you'll need 180 megs of capacity between New York and LA, 200 megs 
between San Francisco and Dallas, etc. It's the "what do I want?" from a 
network, rather than the "how do I get it?". I call this step 1.

Step 2 -- and this is a very different problem to solve, which is why I 
make the distinction -- is figuring out how to engineer the network such 
that it efficiently meets those demands. This *is* the "How do I get what I 
want?" step -- where label switched path computation is performed for an 
MPLS network.

The final step, 3, is to provision the network to implement the LSP layout 
that you computed in the previous step. This is where actual network 
protocols such as LDP, RSVP-TE come in. Call this the "Once I know how to 
get what I want, how do I make it real?" step.

The problem (as I see it) is that we are calling all three of these things 
"traffic engineering". Some software TE tools do step (1) only. Some do (1) 
and (2). Other tools don't try to do (1) at all, but do (2) and (3).

Cheers,

Mathew




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