Yep, I've got a logical grouping set up. Potentially 500 people per room.
I've already split the server up into slaves and a master, so distribution
between multiple machines will me almost mindless. Its also set up so that
everyone can't flock to the same table. 

Thanks for the input. 

The first rule of Fight Club is You Do Not Talk About Fight Club.

// Erik Huizing   huizing@cpsc.ucalgary.ca
// www.cpsc.ucalgary.ca/~huizing

On Fri, 16 Feb 2001, David Brownlee wrote:

> 	Are the players logically grouped - do they group around virtual
> 	tables or rooms? If so, then one thread (or even process) per
> 	room would make sense, with each thread/process syncronising
> 	when players move tables. Seperate processes would even allow you
> 	to distribute it across multiple machines.
> 
> 	Bear in mind sooner or later everyone is going to try to join
> 	the same table Just Because... so plan for that (table limit
> 	is probably easiest).
> 
> 	It may just be easier to go with poll() and have a single list of
> 	all players - bearing in mind you can guarantee round robin
> 	behaviour, which is probably what you want.
> 
> 		David/absolute		-- www.netbsd.org: No hype required --
> 
> 
> On Thu, 15 Feb 2001, Erik Huizing wrote:
> 
> > Right now, I'm bound by non-disclosure agreement not to reveal the
> > specifics, but I can say this: its like an online casino where the players
> > can compete against each other. I need to have potentially a thousand or
> > more simultaneous open sockets at a time.  I origianlly started with a
> > java application, but it coudln't handle the numbers I needed, so I'm
> > rewriting the whole thing in C. What I was thinking of doing was having
> > maybe 10 or 20 threads each listening to 50 to 100 sockets (using an
> > fd_set and select). Is this a good way to go, or is there a more efficient
> > way?
> >
> > The first rule of Fight Club is You Do Not Talk About Fight Club.
> >
> > // Erik Huizing   huizing@cpsc.ucalgary.ca
> > // www.cpsc.ucalgary.ca/~huizing
> >
> > On Tue, 13 Feb 2001, Jeremy C. Reed wrote:
> >
> > > So if I understand you (and this) correctly, it looks like using syslog is
> > > the correct way and using syslog wouldn't be reinventing the wheel.
> > >
> > > Tell us more about your project.
> > >
> > >    Jeremy C. Reed
> > >    http://www.reedmedia.net/
> > >
> > >
> >
> >
> 
>