On Sun, 2005-07-31 at 05:42 -0700, Gabriel wrote: > Querying google for "tcp window size" returns some links. I > don't know the recommended values and I don't know either > if you can play with this value in linux, AFAIK, this value isn't something which QoS can handle or dynamically change. This is based on my research I did some 6-8 months ago. Though things may have changed during that time. > or if it is wise. > There is built-in flow control in the TCP protocol and it > adjusts the window size according to your available > bandwidth. Obviously, the bigger, the better, but if more > data arrives than can be accepted, it will be discarded and > retransmission will be necessary. On the other side, if the > window is set too small, the same information is > transmitted using more segments when maybe it could be > transmitted using fewer segments. Yes, one of the reasons for the need for dynamic scaling of the window sizes is to do reduce the retranmissions and thus the dropping of packets because they arrive too fast. In an ideal world, this shouldn't happen and links shouldn't get saturated with unneeded packets and dropping it at the gateway/router etc and hence losing bandwidth. AFAICT, TCP's built in window scaling just moves it up each time it gets a successful transmit, but when it drops, it will have to restart itself say, from zero. That's my uderstanding. May be flawed. > -- Ow Mun Heng Gentoo/Linux on DELL D600 1.4Ghz 1.5GB RAM 98% Microsoft(tm) Free!! Neuromancer 12:14:36 up 8 days, 18:26, 5 users, load average: 0.15, 0.22, 0.24 _______________________________________________ LARTC mailing list LARTC@xxxxxxxxxxxxxxx http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
