Getting Negative Values for nlost_ and bytes_
Brought to you by:
heidemann,
tom_henderson
Menu
▾
▴
#53 Getting Negative Values for nlost_ and bytes_
I'm using NS to simulate a network with lots of TCP/FTP and UDP/CBR connections. I'm using the LossMonitor and /TCPSink/Sack1 for the sinks.
Using the Lossmonitor, it seems that the nlost_ variable gets some sort of overflow, at the sudden it becomes a negative value, here's an example:
time bytes_ npkts_ nlost_ ... ... ... ... ...
1092.70 44855790 213599 2145574003 504.00 44729600.0 50348800.0 0.768 1
1092.80 44862090 213629 2145985588 504.00 44736000.0 50358400.0 0.768 1
1092.90 44868600 213660 2146397173 520.80 44742400.0 50368000.0 0.768 1
1093.00 44874900 213690 2146808758 504.00 44748800.0 50377600.0 0.768 1
1093.10 44881200 213720 2147220343 504.00 44755200.0 50387200.0 0.768 1
1093.20 44887920 213752 -2147307929 537.60 44761600.0 50396800.0 0.768 1
1093.30 44894010 213781 -2146923784 487.20 44768000.0 50406400.0 0.768 1
1093.40 44900100 213810 -2146567080 487.20 44774400.0 50416000.0 0.768 1
1093.50 44906820 213842 -2146155500 537.60 44780800.0 50425600.0 0.768 1
1093.60 44913120 213872 -2145743915 504.00 44787200.0 50435200.0 0.768 1
Note the fourth colunn, something very strange happens.
I notice the same happened in the TCPSink/Sack1 at the bytes_ variable. Another example:
time cwnd_ ... bytes_ ...
1881.30 220 220.0 2146771570 11.860363636363639
1881.40 220 218.0 2146969380 11.860363636363639
1881.50 221 225.0 2147073570 11.860363636363639
1881.60 221 221.0 2147295540 11.860363636363639
1881.70 222 222.0 2147339330 11.860363636363639
1881.80 222 219.0 -2147347106 11.860363636363639
1881.90 222 222.0 -2147310866 11.860363636363639
1882.00 223 221.0 -2147149296 11.860363636363639
1882.10 223 230.0 -2146992256 -31224.001861818189
1882.20 224 229.0 -2146990746 -31224.001861818189
Again, in the fourth colunn. It seems that after the number gets over 2 bilion 147 milion and something, it turns to negative.
Update from Lloyd Wood on ns-developers list
look at ~ns/tools/loss-monitor.cc
where nlost_ and npkts_ (which is what is really overflowing, I think...) are defined as int, rather than say ulong.
But binding in ns defaults to int, even when unsigned int might make more sense.
xDlaKa Really informative blog article. Really Cool.
Thanks a lot for the blog.Really looking forward to read more. Much obliged.
I appreciate you sharing this blog post.Really looking forward to read more. Really Cool.
Thanks again for the article post.Much thanks again. Great.
I cannot thank you enough for the blog article.Thanks Again. Fantastic.
I appreciate you sharing this article.Really looking forward to read more. Much obliged.