The MPLS WG Archive

[Vishal Sharma <Vishal.Sharma@tellabs.com>]

Ramprasad,

It is unlikely that anyone will be able to give you a ready-made simulation scenario. I think
Florian made some good suggestions on the readings.

I think what is missing in your plan is a clearer objective. You sort of stated
it when you said :
>  want to show how MPLS is
> advantageous over normal IP in as many aspects as I can, using my MPLS
> simulation (using ns).
(I think it still might be useful to ask why you want to do this, and what
is the final purpose of doing this exercise.)

Assuming this is your aim, you have to ask yourself (a) in what respects is MPLS
stated to be better than IP and (b) what would be a way to design a simulation
to demonstrate that.

As Brijesh mentioned in an earlier email, a worthwhile comparison is really
to examine what MPLS provides in terms of traffic engineering capabilities
relative to either pure IP routing, or IP routing augmented with some
traffic engineering capabilities in the interior gateway protocols (IGP), such as
equal-cost multipath, or metric tuning.

If this is the aspect that one believes MPLS could be better in, then the following
could be one simple way to go about it.

So let's say that at a first-level you assume the following:
 (i) a network topology (the nodes, their locations) and parameters (the link capacities, and 
metrics).
 (ii) a traffic matrix for the network (that is, static demands; a simplistic scenario, but 
probably
a good starting point, especially if you consider backbone networks that carry aggregated
demands that are likely to be fairly constant [they might have some variation about a mean
value, but one could assume for starters that the variation is small]).

The aim of the comparison then could be to comapre the following (Brijesh already pointed
this out)
(a)  The delay experienced by the traffic, with and without MPLS traffic engineering.
(b)  The total amount of traffic that could be served, with and without MPLS traffic engineering.
(Given any traffic matrix, you may not be able to serve all of it, so you'd have to see
how much of the demand you can serve in the two cases.)

(We will ignore forwarding performance, since as people have pointed out, that is
 implementation dependent, and is not really a differetiator for MPLS anymore.)

Please note, that to do this your simulation will, in addition, have to run some algorithm(s)
that can decide exactly which MPLS LSPs to set up, and how much bandwidth to assign
to each of them to serve the traffic demands specified in your traffic matrix.

So in effect, what you will really be testing is the "goodness" of these LSP placement
algorithms, because the better these algorithms are in finding "good" paths through
the network (assuming that these paths can then be setup using MPLS signaling protocols), the
better is likely to be the performance of the MPLS-based scheme
relative to IP routing.

-Vishal

On Thursday, May 25, 2000 4:58 AM, ramp@sasi.com [SMTP:ramp@sasi.com] wrote:
> Hi,
>
>
> Can U suggest an example scenario of what U mentioned here with all the
> details of link bandwidths ,packet sizes, Type of packets ...etc. As I am
> new to this kind of analysis, I don't have any realistic scenario at
> present.
>
> Thank U very much
> B G Ramprasad Torati
>  On Thu, 25 May 100, Khaled Elsayed wrote:
>
> >
> > As other people said, it is not surprising you would get the same results. You would need to
> > come up with a scenario where if you use tradional routing versus proper traffic engineering as
> > in MPLS, the difference in the peformance will be clear. However, your scenario should be as
> > realistic as possible, i.e. it can happen in an IP-based network. This is the difficult part. 
The
> > rest should be easy.
> >
> > BR,
> >
> > Khaled Elsayed
> >
> >
> >
> > ---------- Original Message ----------------------------------
> > From: B G Ramprasad Torati <ramp_mpls@yahoo.com>
> > Date: Tue, 23 May 2000 21:53:47 -0700 (PDT)
> >
> > >Hi
> > >
> > >I simulated the LDP and MPLS packet switching parts of
> > >MPLS by extending the network simulator 2.
> > >Now I want to do the performance analysis between the
> > >two forwarding paradigms( MPLS forwarding and normal
> > >IP forwarding).
> > >
> > >For this I have taken an example topology and did the
> > >following things.
> > >1. Using CBR agent I calculated the delays by using
> > >loss monitor as the sink of the CBR packets.
> > >2. Using Ping agents I calculated the round trip time.
> > >
> > >I did those using both normal IP and MPLS with LDP.
> > >But I didn't get any difference between those two
> > >forwarding paradigms. I calculated the no. of packets
> > >lost using CBR agents. Still I got the same
> > >performance for both the above. I used both Link state
> > >and Distance Vector dynamic roting protocols while
> > >doing this analysis.
> > >
> > >So can anybody suggest how to do the performance
> > >analysis. I am new to this, so please tell how exactly
> > >I have to proceed in doing this analysis.
> > >
> > >Thanks and Regards,
> > >Ramp
> > >
> > >__________________________________________________
> > >Do You Yahoo!?
> > >Send instant messages & get email alerts with Yahoo! Messenger.
> > >http://im.yahoo.com/
> > >
> > >
> >
>