gionnico wrote: > The bandwidth that gtk-gnutella measures isn't real. > If I sum http+gnutella+leaves I don't get the usage that I can measure > with a program like iftop, ifstat or conky (I don't know if conky uses > another program, though). > It turns out to be about 30-50% of the real used bandwidth. > And it's not about tcp-acks: the problem is the same even when I'm not > downloading, so that the acknowledgements for the same gnutella network > are negligible.
I think in ultrapeer mode the measurements get increasingly off with the amount of leaves. I assume the difference is caused by TCP-ACKs though. Why do you think it isn't? For file transfers in either direction TCP-ACKs should be mostly negligible because the payload outweighs the control traffic. For Gnutella traffic, this isn't necessarily true because many leaves may just send/receive an occassional packet with little payload and the TCP-ACK is comparatively large and may not be piggy-backed either. > Parentethical I don't know but I think that acks should be included anyways. The problem is, that happens on layer to which we have almost no access and all interfaces which exists to those layers typically require super-user privileges. In theory, most of your traffic might consist of retransmitted segments and ACKs but on our layer, we see none of that, we only see the effective payload and its transfer rates. -- 1000 octets = 1 ko = 1 kilooctet; 1024 octets = 1 Kio = 1 kibioctet 1000^2 octets = 1 Mo = 1 megaoctet; 1024^2 octets = 1 Mio = 1 mebioctet 1000^3 octets = 1 Go = 1 gigaoctet; 1024^3 octets = 1 Gio = 1 gibioctet ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ gtk-gnutella-devel mailing list gtk-gnutella-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/gtk-gnutella-devel
