[MUD-Dev] Re: Ethernet NICS, maximum connections..mud testing.

[Chris Gray]

[Ben Greear:]

 >> They vary. There is a fixed header of 12 bytes, optionally followed
 >> by other data, perhaps upto 1 or 2K in the extreme. Most will be
 >> quite short during the 'build the world' phase, which is where my
 >> concern is.
 >
 >The larger the packet, the more efficient it will be (generally),
 >up to around 8k.  A bunch of small packets could eat you alive!
 >
 >Might want to consider packing several requets/messages into a larger
 >buffer before writing to the socket.

I understand that. Not much I can do about it, however. Luckily, that
aspect of my communication is really only true when building the
scenario. During normal operation, packets will consist of whatever
text the user types in, or whatever text the server needs to send to
the user.

The small packets don't explain why I'm seeing a mostly idle system
while doing the world-build, however, and that's what I'm mostly
concerned with at the moment.

I've been running some stability tests today, and it looks like,
socket communication ignored, I could handle 300 or so users, perhaps
more. (I'll have to try again when I compile things with optimization
on - that might make a fairly big difference.) I expect socket
communications will be the bottleneck.

Anyone out there still able to run my old Amiga client for an
internet connection? I think I'm now ready to be able to accept
connections from one, but I'm not in a position to test it myself.

--
Chris Gray     cg at ami-cg.GraySage.Edmonton.AB.CA