Armin,

the page sizes don't depend of the TCP version. You can see that at the
begining of the simulation, the page sizes are equal, but after a few
seconds, maybe, the size pages are different.

why?.

the NS-2 web traffic module models the web traffic interaction as realistics
as it's possible, and it's modeled such Statistical distributions, and some
of them depend of the time.

So, if you use different TCP protocols (for example, RENO) , with different
timers, different handle of packets, maybe, some packets will arrive at web
server at different times that if you use the other TCP protocol (for
example, CUBIC). So, it's produce different web server responses (maybe with
different size), and this page will produce different behaviour at client
node, and so on...

Conclussion: it's imposible to compare the same trace with both
protocols..... at least... I don't know

best regards

salva

2009/1/26 Abfalterer A. <[email protected]>

>
> Hi,
>
> I've made a Web traffic simulation similar to this from
> "ns-2.33/tcl/ex/web-traffic.tcl". I take the same parameters (e.g. for
> page inter-arrival time, object number, etc.) and I want to test it with
> different protocols (TCP NewReno and CUBIC).
>
> First test runs show that the page sizes are not the same for
> *different* protocols. For instance, a test run with one session and 10
> pages with NewReno gives me pages of following size (in packets!!!)
>
> NewReno         CUBIC
> 170             170
> 1               1
> 9               9
> 1               1
> 2               2
> 4               4
> 1               5  <-- different
> 5               1  <--
> 2               9  <--
> 10              1  <--
>
> Anybody who has experience with Web traffic simulations and can tell me
> the reason? Unfortunatley there is not a lot of literature...
>
> Regards, Armin
>
>

Reply via email to