[MUD-Dev] New to MUD Dev, need friendly advice!

Matthew D. Fuller fullermd at over-yonder.net
Tue Apr 29 13:06:06 CEST 2003


On Mon, Apr 28, 2003 at 12:59:11AM -0400 I heard the voice of
Mike Shaver, and lo! it spake thus:
 
> That's likely different in the MU* space, though, where
> connections will tend to live many minutes rather than a few dozen
> microseconds.  In that case, doing whatever's easiest to reason
> about (fork for each client, process pool, thread pool, thread
> creation; "easiest to reason" is very much a personal style issue)
> is likely the winning approach, because the cost of the
> fork/thread_create will be below the noise floor.

In terms of creation, of course.  In terms of usage, not so much; a
few thousand processes or a few thousand threads are all going to
have troubles.  You've got a bunch of contention domains for any
shared data, you've got all the gritty details of OS-level
scheduling dealing with so many entities, resource usage, yada yada.

We've had this discussion before (as has been mentioned), and it
generally comes down to something along the lines of "A new and
seperate X for every connection is a losing strategy".  A process
(or two or three) devoted to handling all the connections, or a
thread (or two or three) for handling all the connections; these
tend to be the last men standing.

--
Matthew Fuller     (MF4839)   |  fullermd at over-yonder.net
Systems/Network Administrator |  http://www.over-yonder.net/~fullermd/
       "Klein bottle for sale ... inquire within."

_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev



More information about the mud-dev-archive mailing list