[e2e] variable bandwidth link

Alhussein Abouzeid hussein at ee.washington.edu
Tue Apr 17 17:43:16 PDT 2001

Error models for wireless links have been studied quite elaborately (e.g.
search for a recent survey by M. Zorzi). Since  you want to model TCP over
such a link, you need to first pick one of these models, then find out how
much delay/capacity will be involved in hiding the errors caused by it.
This is the hard part. If you overcome this obstacle, you've arrived at your

Of course, the practicality of this variable bandwidth model depends on
the practicality of your assumption about hiding wireless errors
from TCP, which may be quite practical in some situations (and not so in
many others). I think it's a good approach though.

Hope this helps,


On Wed, 18 Apr 2001, Soo-hyeong Lee wrote:

> Hello,
> I am trying to model a wireless link empowered with link-level recovery as an error-free link with variable bandwidth.
> I suppose link-level loss recovery (such as snoop or combination of FEC+RLP) can completely hide link-specific errors.
> Dupacks are assumed to be either eliminated at the wireless link (by snoop) or ignored at the TCP sender (by TCP-Eiffel). The only way that a link-level error can be seen by upper-layer is by link-level timeout (in the snoop), which is assumed to be very rare.
> IMHO, this link will be seen to upper-layer as an error-free variable bandwidth link, but unfortunately I find it difficult how to actually model it.
> I would be very grateful if one of you kindly give me a piece of advice.
> Thanks in advance.
> Soo-hyeong
> PS : Intuitively, any shared medium access link (including Ethernet) could be modeled as a variable bandwidth (or variable delay) link. Is this wrong, or just hard to do?

More information about the end2end-interest mailing list