Messages in this thread | | | Date | Sun, 25 Apr 1999 12:12:00 -0700 (PDT) | From | Cacophonix Gaul <> | Subject | Initial snd_cwnd in pre-patch-2.2.7-3 (net/ipv4/tcp_ipv4.c) |
| |
I'm really glad that snd_cwnd is being initialized to greater than 1 (even though the perf problems with certain clients are not the fault of linux) :-)
However, why not go all the way up to the limits permitted by rfc 2414 (ftp://ftp.isi.edu/in-notes/rfc2414.txt)? Specifically, the rfc suggests that the upper bound on initial window should be:
min (4*MSS, max (2*MSS, 4380 bytes))
Of course, this violates rfc 2581, which requires cwnd <= 2*MSS, but notes that rfc2414 permits this as an experimental extension.
Thus, if initial MSS=1460, then we are allowed an initial cwnd of 3.
Why do I think this is a big deal? Because, consider the case of large web servers (or certain web benchmarks ;-) that get a large number of requests for small files - _most_ of these are under 4380 bytes. In cases such as these, it seems to make sense to be able to transmit the entire file in 1 rtt.
In any case, I'm pretty happy with an initial snd_cwnd of 2, and will play around with 3 to see if the difference is significant.
cheers, karthik
_________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |