cvs commit: src/sys/netinet in_pcb.c tcp_subr.c tcp_timer.c tcp_var.h

[Mike Silbersack]

Ok, I started looking through the mess that is in_pcb.c, and I came up 
with a simpler idea than trying to improve upon my old heuristic.

What if we just build upon what Gleb did in revision 1.256, and change the 
size of the tcptw zone?  Instead of scaling it to maxsockets / 5, let's 
scale it to max((ipport_lastauto - ipport_firstauto)/2, 500).  We'll have 
to rescale it whenever the port ranges are changed, but those sysctls are 
already handled by a function, so it'll be easy.

This means that we'll be keeping around fewer time_wait sockets than we do 
at present, but I don't think that's a big problem for anyone.  On the 
positive side, it means that time_wait sockets can't starve out ephemeral 
ports unless you have more than 50% active connections.

One slightly more complex solution would be to use one tcptw bucket for 
connections with local ports >= 1024 and a seperate bucket for connections 
with local ports < 1024.  Assuming that our front end web proxy answers on 
ports < 1024, that would ensure that we keep one pool of time_wait sockets 
for our connections from clients and another pool for our connections to 
the backend web servers.  I guess that would be slightly more "correct".

What do you guys think?

Mike "Silby" Silbersack