Evaluate Windows TCP algorithm against actual Windows implementation

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Evaluate Windows TCP algorithm against actual Windows implementation

Souhard Kataria
We have implemented a TCP algorithm that is a part of the Windows TCP stack
in ns3.
How do we test/compare our code with the actual Windows implementation?
Is NSC the best way to go about it or are there alternative newer methods?

Regards,
Souhard Kataria
Reply | Threaded
Open this post in threaded view
|

Re: Evaluate Windows TCP algorithm against actual Windows implementation

Tom Henderson-2
On 11/8/19 9:13 AM, Souhard Kataria wrote:
> We have implemented a TCP algorithm that is a part of the Windows TCP stack
> in ns3.
> How do we test/compare our code with the actual Windows implementation?
> Is NSC the best way to go about it or are there alternative newer methods?
>
> Regards,
> Souhard Kataria

I do not think that NSC will be of any help to you.  We use DCE nowadays
instead of NSC, but DCE also is not for Windows stacks.

By Windows TCP, do you mean Compound TCP?

The most straightforward way to check, if you have access to testbed
resources, is to create packet traces on a real testbed by subjecting
the Windows implementation to a set of impairments (packet losses,
artificially-induced delays) and then to recreate in simulation and
compare the traces.

Please see this past post about another implementation; perhaps you can
cross-check your implementation against theirs:

https://mailman.isi.edu/pipermail/ns-developers/2018-July/014456.html

- Tom