I can''t see why a MMORPG would needs more than 1kb/sec of data to be sent. It''s not like you need time critical events. With aggressive culling, compression, and a pure event driven networking model, it should be more than sufficent. Just dont go peer 2 peer for MMORPG
This was discussed long ago aswell.
I think the consenus was :
action games consume about 2-5kb/sec
MMORPG consume 100bytes-1kb/sec
Turn base games consume very little bandwidth
For all networking games, the more players you can fit into your bandwidth, the better. Though making a game which draws enough players to fill that bandwidth, is much harder than optimizing for bandwidth, I''ve found
Tombstone takes ~74bytes per second per character per character. A typical packet is around 10bytes or less. Plus header info courtesy of DPlay. That''s an additional 20bytes per packet or so.
A standard 256k DSL line can handle 32768Bytes per second being transfered. That''s around 20 people trasmitting data at once per second.
Basically you just need to work on a LAN and get the packets down to a minimum using software to monitor network traffic. I use McAfee FireWall.
As for recouping cost, banner ads blow. I use Commission Junction.com on my web-site and in 3 months have had ~40,000 impressions. I''ve made $0.
The pay for play model is about the only way to do it. I have a DSL line running my .com so running MMORPGs technically doesn''t cost me anything.
But yeah, just optimize optimize optimize your network code and when you go live start with a standard DSL and when you need to, upgrade.
I would say search my posts, but that would end up with a lot more info than you need, so I''ll pare it down.
MMOG users use ~1k/sec. You can say 100B-4kB/sec, but when you''re looking at MMO, you''re looking at the macro (all users) scale, not the micro (single user). The average is 1kB/sec.
20% of your total users are online at any ''peak'' time, capping out to 25% during special days (holidays, days immediately following significant updates/expansions). So you figure that you''re looking at for every 100k users you should be capable of supporting 25k users online, or around 25MB/sec, which equals out to 200Mb/sec. This where it starts to get a *little* expensive. But you''re still only looking at 10s of thousands of $$ for the pipe, mattering on where you locate.
So you figure that you''re getting ~$1,000,000/month income and you''re spending around $100k/month on bandwidth. Seems great, doesn''t it?
Well, you''ve only touched the tip of that proverbial iceberg so far. Now you have to earn enough to pay for those developers, so say that the initial purchase price of the game pays for that. Those GMs and CS personnel don''t do the job for the love of the game, they get paid. Not highly, but they do get paid and you''re probably looking at 4 per shard, or 4 per 10k users (mattering on how you want to calculate it), earning around $30k/year average. Double that after you add in taxes and benefits and you''re spending, after adding in all users for all shards, around $200,000/month in salaries for the GMs and CS personnel.
Now you toss in developers. Probably around 10 for the game to fix bugs, implement changes, etc... They cost around $40k/month minimum, double that for taxes, benefits, etc.. and you end up paying around $65k/month for those people. Of course, you need administrative people around the clock, so you need to push another 6 people onto the staff making around $70k/year because they have to be experts in the OS and they have to know in an instant what went wrong in case of a system failure. Double it again, you''re paying these people more than just health benefits and standard SS taxes... there''s a 401k that they''re putting money into and so are you. You''re paying them around $75k/month.
So what are we up to? Call the bandwidth $75k/month, CS/GMs $200k/month, Developers $65k/month, Systems Admins another $75k/month. That comes up to $415k/month. Not bad, you''re only paying out 41% of your current income.
Unfortunately you''re not done though. You have to buy hardware for all those shards. Using EQ as an example, you will have to purchase a Dual processor, top of the line system for about every 2 ''zones'', or if you avoid the ''zone'' philosophy for an alternative, about 50 machines for every shard. So you''re spending about $6k/machine and it adds up to around $300k/shard. Balance that out across 2 years and figure that every shard supports 10k users and you''re spending $3Million, or around $125k/month. Every time you make your world larger, it will only increase this cost.
Of course, you have to build an infrastructure for these machines and you don''t want to use that crappy Cicso stuff... that''s good for small and medium sized businesses that transfer a few Mb/sec, but you''re sending 100s of Mb/sec, so you have to go get some serious equipment from Foundry Networks. So you''ve got to get an OC12 or larger router, a serious switching unit so that you''re not lagging your users out, fibre infrastructure, etc, etc, etc... now you''re looking at even more money that you''re spending at the time of startup.
My point is, that while it may sound like you''re not spending much on your game, reality is that after you add in the personnel, equipment, software, etc... it gets expensive to run one of these things. You''re probably running a year before anything which resembles a profit comes in.
By that time you''re looking at the next thing you have to do because you don''t want this game to be a one-off. You want to continue developing this type of game because the potential is amazing.
But as for the EQ thing: Yeah, EQ is a cash cow of amazing proportions. It can fund 3-5 games easily. Unfortunately after you spend $2Million here for this failed project (like Tanarus) and every other mistake that your company makes (how many games have you seen by Verant/SOE that are on the CDs that basically nobody plays?), it adds up quickly to mean that you don''t just want to have that cash cow, you NEED it to keep trying to hit the lottery again.
I''m pretty sure that Brad McQuaid left Verant because they are basically trying to create another ''cash cow'' instead of trying to make another great game (like EQ was when it came out... a genre defining game). Creativity? No, let''s see how we can minimize investment and maximize return. They used to try and make things for the players because the players wanted them. They are now trying things that get extra money out of the players. Legends? I have a feeling that SWOL isn''t going as well as some would expect because they are trying to build something from the ground up again and running into many problems that they never ran into with EQ. You think that Lucas is paying for the game? Verant is probably paying for the development and paying for the right to even make the game. If they screw the pooch, then it''s just money out of their pockets, Lucas can just hit them with a breach of contract and have someone else do it.
Man, this has turned into a long post. Sorry for the length.
Just remember, this is the type of game that ends up with the company leaning on it too much. You stop hunting for publishers to fund your next project, you end up funding it yourself and I think that what that ends up meaning is that you''re spending too much money taking chances on games that won''t necessarily succeed, but that ''sound cool''. Just look at the failed Verant/SOE projects.
Damn, in all that, I forgot to answer the question... hehehe.
The concern isn''t at the server/shard end, it''s at the client end. You need to make it where your players on a 28.8kbps modem can play the game and have few problems. If you lose that group of players, you are losing a decent amount of the market. And before you say "screw them if they can''t buy a 56k modem", don''t think that they don''t. I know a large number of people in suburban, affluent areas (Rochester Hills, where the average income is well over $100k/year) where if you connect at 33.6, it means that you have a Courier v.Everything AND got lucky. Cable modem? Not available. DSL? Not available either.
The bandwidth cost is significant, but it''s not enough to where you''ll cry yourself to sleep at night over having to spend an extra 20% because you''re sending 1.3kB/sec instead of 1kB/sec (remember, economy of scale means that as you increase your bandwidth, you spend less on the additional bandwidth).
I think most companies are really over paying on bandwidth. If you colocate your hardware you will be able to get your average bandwidth costs into the $0.25 - $0.40 range per 1 KB/s. These prices reflect high quality bandwidth to large nationwide providers. The $0.40 price would be a tier 1 type of provider. This won''t include rack space which will run around $700 per rack per month. With space efficient hardware you could get at least 16 servers per rack (assume 2 U servers, 40 U racks). With 25,000 peak players and four racks of hardware your MRC is going to be around $12,000.
I don''t work for a gaming comany, but we push 1300+ Mb/s of real-time streaming traffic during peak everyday.
