[MUD-Dev2] [OFF-TOPIC] A rant against Vanguard reviews and rants

[Amanda Walker]

On Mar 7, 2007, at 2:12 PM, Adam Martin wrote:
>> The graphics engines aren't the problem, I don't think   Torque has
>> been out for years, for example, and even has a 3rd party "MMO Kit".
>> Irrlicht and CrystalSpace are free, etc.
>
> ...and there's much much better stuff out there for very low cost if
> you're going to go commercial, or have a promising project and can get
> the interest of the big commercial engine licensors.

Indeed.  I was just pointing out that engine availability isn't the  
stumbling block, even if you have no money at all :-).

>> The problem I see for doing an indie MMO is infrastructure.  It takes
>> a lot of bandwidth and computing power to support very many users,
>
> Nah, b/w is ever cheaper and computing power now for servers is
> shockingly cheap - prices I was quoted last year for quad-core servers
> with more RAM than any of our dev machines are lower than I was
> expecting for single-core 1Gb machines.

Bandwidth is ever cheaper, yes (though dedicated low-latency  
bandwidth is less so than "good enough for a web site" bandwidth).   
Power, alas, is not.

Let's back-of-the-envelope what it might take to roll out an indie  
MMO (as opposed to an indie MO, which are legion :-)), since I  
suspect we're comparing apples and oranges a bit.

Baseline assumptions:
- Use BitTorrent, sourceforge, googlepages, or some other free file  
hosting service to host client downloads.  No cost there, pretty  
reliable.  $0.

- Static terrain/buildings/etc.: prevents having to store & stream  
geometry/texture data, and means you don't have to run (much) physics  
on the server beyond passable/impassable areas.  With a good  
architecture, this means you could get a few thousand clients per  
server, rather then a few tens to hundreds (note SL, where running  
lots of stuff server side and streaming everything to the client  
means that they can only support a couple dozen clients per server  
before they hit performance walls).
- Let's say 2000 clients per server, and disregard for the moment  
whether that's sharded vs. a single load balanced simulation.
- Say half a rack per data center: 6 servers, Ethernet switch(es),  
router, a couple upstream bandwidth providers (in case one has bad  
internet weather).
- Start with, say 3 data centers; say, SF, DC, and Chicago.  Since  
we're an indie, using data centers with "smart hands" service means  
you don't have to travel as much to deal with stupid hardware problems.

So, let's see.

3 data centers x 6 servers per data center = 18 servers, figure we go  
midrange for about $2500 each.  That's $45K upfront capex (though you  
could start smaller if you don't expect tens of thousands of users).

A half-rack of space, power, and a few megabits of bandwidth will  
probably be about $1K/month, x 3 = $3K/mo opex.

Conclusion: Very practical for a startup if you have some money to  
throw at it, but not practical as a sideline even on an affluent  
programmer day-job salary.

I suspect that the future of indie MMO hosting is not in Moore's Law  
as applied to server prices, as impressive as that is, but in things  
like Amazon EC2, where you can buy both computing power and bandwidth  
in much smaller incremements based on demand:

http://aws.amazon.com/ec2

A virtual world toolkit based on something like that would be very,  
very interesting, I think.

> have friends who've
> kickstarted stuff on what a few years ago would have been considered
> major "big iron", funded by nothing more than their own spare cash
> (normal people on average programmer salaries).
>
> For instance, have a look at people like OGSI.

OGSI, if their website is accurate, is hardly funded by spare  
cash :-).  It does like like they've got the "indie games  
infrastructure == market opportunity" idea, though.  Looks pretty  
good (and reading their slide show, looks like they're making the  
same point about "rolling your own" infrastructure as I did above  
with my envelope costing exercise).

> Personally, I'm betting that will happen but disappear because the
> amount of revenue being pumped into client/server systems vastly
> eclipses what goes into the p2p stuff, and yet the p2p stuff has only
> marginal benefits.

I'd have thought so too (my day job is definitely in the "bet on big  
iron" category ;-)), but I am continually surprised that it keeps  
popping back up.  Skype, BitTorrent, Joost, ... there's got to be  
some benefit or people wouldn't keep doing it.

> This will remain the case until/unless the
> last-mile home broadband b/w goes up by several orders of magnitude
> whilst the cost of server b/w and silicon only drops by a small
> factor.

Well, I wish we had reasonable home broadband in the US, but we won't  
as long as the Bell System and the cable TV cartel keep their  
regulatory monopoly on the last mile.  This is one area where Europe  
and Asia are pulling way ahead.


Amanda Walker