Im surprised your bandwidth usage is so low. Im working on a multiplayer action game and were averaging 1.5 KB/sec per client connected. It seems there are orders of magnitude differences in bandwidth consumption between the 2 games. Im thinking about making a simple SMORPG someday. Thanks for the info.

-ddn

I''m using quite a few tricks to pull it off. Compression and scripting are basically the keys.

Ben

To solve our bandwidth issues, we''ve been working on just sending the deltas (changes to the status quo) and the information in the local area that affects the client. If you have a large number of other characters in an area not doing anything, why update the client? If another entity is running in a straight line, only tell the client when that behavior changes.

If you have a good method for handling lag (dead reckoning, etc...), you can use this to your advantage. Since the dreaded lag beast is the biggest danger to us all, use your lag handling code to your advantage. Anything you can do to reduce the ammount of info going to and from the client will help in other areas.

On a side note, you can set additional machines on your network to simulate lag. One method is to have two "chatty cathy" machines that send huge ammounts of data back and forth. The other way is to have a "slow proxy" server that you send all communication through. Just write some code that will forward all code to the server/client but with a delay. That way you can simulate any sort of lag you want. You can also use this to simulate different connection speeds as well. Saves a lot of time setting up modems and such.

I think I should make this post as it brushes into the OS realm. It is important to note that your programmers must make the decision on the operating system, as it''s really their call. Sockets are VERY different in each operating system. In my opinion Berkeley Sockets (most *nux systems, including Linux) offer the easiast learning curve. Honestly the WSock32 API is very wierd too me, but again thats my opinion. *nux systems also offer support beowulf, which allows splitting the workload among more then one computer. It might be a good idea to split everything among cheap servers. Like a database server, running PostgreSQL or MySQL could easily and quickly be accessed via a 100 mbps switch by a world server. The database backend should also be connected to the web server. It is my belief to split each function into a server, thats what im doing for my MMORPG, but then again its more of a mud with simple graphics and doesn''t require that much processing power. You could split the world up into servers, each area living on a different server. Whatever you decide make it flexible, so you don''t get trapped into one method. Experiment. Get a few programmers trying different things on simple 486''s with stupid chat programs, something to buy time while the story writers are churning out ideas. Not every method will work for one project, its important that you understand your goals and the server load. Don''t grow an ego eaither. If your MMORPG grows so popular that it needs to support 10,000 people ill be shocked. Think simple then grow bigger. Set a smaller goal like maybe 100-200 clients at once. That on its own is pretty lofty. You might also decide to break the world down into smaller areas, split among servers. I know EQ does this, and AC does this but in a different and less obtrussive way. Again it all depends on your project. I know that didn''t answer any of your questions directly, but I hope that from this you will be able to draw your own conclussion.

New article on gamasutra talks about serverless massive peer 2 peer multiplayer games. Its intresting, a different design approach to the issues of MMORPGs. Personally i think it will go more toward the micro servers concept. Micro-Servers which dynamiclly load balance and migrate between machine, to take advantage of the changing network archictecture as players come and go, each sharaing the client load.

http://www.gamasutra.com/features/20001229/fitch_01.htm

Good Luck

-ddn

If you have 10,000 players online at once, 24/7, expect a $50,000 bill purely from your bandwidth useage.

If you have 10,000 players online at once, 24/7, expect a $50,000 MONTHLY bill purely from your bandwidth useage.

Asheron''s call used clusters of 7 servers. 4 of the servers would distribute map files to people that needed them (you download the map as you walk into it, since it is 490+ square miles ). The other 3 were for AI and collision checked, et cetera. Each one of these clusters costed about $700,000. One of the developers said something about it online one day .

A T1 costs about $1000 a month, but you may not need only 1 T1 =P.


http://www.gdarchive.net/druidgames/

10,000 players on 24/7 of a player base 4x larger say, 40,000 paying 10 us dollars a month = 400,000 us dollars. Hmmm 50,000 dollars is only 12.5% of that. Even with maintence cost thrown in, it''s still a heft profit, but it has to last long enough to make back the cost of developement, which at a budget of 2 million, say within a year it should do. That''s if you can get a player base of 40k sustained for a year. Ultimate Online last i heard had a player base of 80k, and Everquest within 100k.

Its not surprising so many game companies are throwing their hand into MMORPGs, its one of the most profitable entertainment bussiness out there, and the most addictive game genre.

-ddn

EQs @ 300k now (worldwide)
Yep, $3,000,000 gross rev. a month.

Magmai Kai Holmlor
- The disgruntled & disillusioned