[ns] Reno showing poorer performance than Tahoe

[Ahmed Khurshid]

Dear all,

I have created a simulation scenario where two TCP and one UDP connection,
residing in different nodes, are sharing the same link and have the same
node as the destination. I have introduced an error module on the shared
link to observe the behavior of the two TCP connections. I found that if I
increase the error rate (more than 0.05) then the maximum transmitted
sequence number of the Tahoe connection always remain higher than the same
of the Reno connection. But we know that Reno is less conservative as it
employs fast retransmission and hence should have a better throughput than
Tahoe. This seems quite unusual to me. Can anyone please explain the actual
reason behind this phenomenon? The tcl code has been attached with this
mail.

Regards,

Khurshid