Re: [ns] realistic web workload in NS?

[kunchanl <kunchanl@pollux.usc.edu>]

> Ns has a simple HTTP/1.0 web traffic model, please refer to
> ~ns/tcl/ex/web-traffic.tcl for details.
>
> We do not have SURGE integrated into ns yet.
>
On the other hand, in terms of the structure of the model,
the ns model is quite close to SURGE.

We have also been working on a model generation tool that
generates web traffic and ftp traffic from tcpdump trace.

Basically our approach is taking tcpdump trace as
input to retreive the statistics about
user behavior and topology information of traffic,
and use these statistics (in CDF form) as input
for a ns simulation model.

For web traffic, the statistics of user behavior will
include things like user arrival, page size, page arrival...etc.
The statistics of topology will include estimation of
bottleneck bandwidth and RTT between servesr and clients.

This work is still in progress, but if you're interested, please
take a look at ~ns/indep-util/model-gen/ for
the model generation tool and the corresponding simulation model in
~ns/empweb and ~ns/tcl/ex/empweb

They are currently only can be retrieved from the snapshot.

Kun-chan Lan

> Cheers,
> -chen
>
> On Thu, 3 Jan 2002, Soo-hyeong Lee wrote:
>
> > Dear sir,
> >
> > Do you know what is the best realistic web workload?
> > I have heard of Barford's SURGE, however it doesn't seem to be available in NS.
> >
> > Regards,
> >
> > Soo-hyeong
> >
> >
>
> --
> Xuan Chen
> USC/ISI
>