Cell Relay Archive

[goyal@way.cis.ohio-state.edu (rohit goyal)]

I agree. Even though Opnet is useful for small/short simulations,
it is not suitable for longer ones. I have used it for ABR and UBR
congestion control simulations, and although the model is very easy
to implement, the simulations take a lot of time and memory. Our 
machines have at least 128MB RAM, but simulations with 5 or 10 TCP
sources running over 155.52Mbps ATM with 10-20 ms RTT take several hours 
to run. We tried to run some satellite simulations, but they didn't
finish for over a day, with a lot of disk thrashing...

As far as I know, the basic ATM model hasn't changed from 2.5B to 3.0.
Although I haven't looked at 3.0 carefully yet, I'm pretty sure
that ABR is still not implemented. Rel. 2.5 had some bugs in the 
Segmentation and reassembly modules, which they claim to have fixed
in this release.

On the other hand, the package is pretty easy to use with a C like
programming language, an excellent GUI, and pretty decent documentation.
The whole EMA concept still baffles me though, but there are ways
to get around it :).

-Rohit
--
Rohit Goyal, CIS PhD Student            (W) (614)-688-4482
395 Dreese Lab, 2015 Neil Ave           (H) (614)-488-5785
The Ohio State University               Fax: (614)-292-2911
Columbus OH 43210                       goyal@cis.ohio-state.edu
http://www.cis.ohio-state.edu/~goyal


In article <32EBDD8B.1FCD@net.com>, James Scott  <jscott@net.com> wrote:
>Iftekhar Hussain wrote:
>
>> Does anybody has simulated the PNNI routing protocol (preferably on
>> the OPNET package)? Thanks.
>
>caveat: i use the OpNET tool to simulate various CAC methods and
>        proposed congestion/buffer mgmt techniques, rather than
>        routing issues
>
>while OpNET is certainly a very nice tool, i think you'll find its
>atm model to be a bit light for simulating PNNI.  this is largely
>due to:
>
>i)  VP connections cannot span multiple nodes and are statically
>    allocated prior to the simulation run
>
>ii) for network simulations with large numbers of connections, you'll
>    find that the host memory requirements are non-trivial
>
>its possible that mil3 has addressed item 1 above in its latest
>release (3.0).  i think item 2 is more a function of the OOP
>nature of the tool.
>
>the end result is that, for simulating PNNI using OpNET, you'll
>likely have to develop your own atm library ... which is certainly
>achievable, but may add another couple of months to your project
>schedule.
>
>cheers,
>
>james