Re: [GNUnet-developers] p2p 'massively multi-player games (MMGs)'

[Chris GauthierDickey]

Hello,

  Sorry for the late response here...just joined the list recently  
and wanted
to make a few points:
> > I'm currently evaluating the possibility of using p2p networks for  
> > the use
> > in
> > 'massively multi-player games (MMGs)'.
> >
> > The requirements are quite different compared to a filesharing  
> > application,
> > namely:
> >
> > *Performance (Latency)
> >
> > *Availability
> >
> > *Security (Account thefts)
> >
> > *Scalability (popular games have up to 180k players in one night)

First, the UPenn paper you cite incorrectly suggests that MMOs
concurrently have up to 180k players in a single virtual world. This  
isn't
quite true. MMOs are split into different servers, each running  
identical
copies of the virtual world, albeit with different player and item  
databases.
This is similar to multiple Counterstrike servers running. As far as  
I know,
Eve Online is the only commercial MMO to try to put all the players on
one server, and currently has a record of around 20k players. Not that
it's not a lot of players, but an order of magnitude smaller than  
180k (=.

> > (Source: Department of Computer and Information Science,  
> > University of
> > Pennsylvania - http://www.ieee-infocom.org/2004/Papers/03_2.PDF ).
> >
> > While there was some research work done round the globe (i.e.  
> > China: Network
> > Security Lab, RIIT Tsinghua University -
> > http://security.riit.tsinghua.edu.cn/Freegame/English%20Vesion/ 
> > freegame.html
> > ,
> > Germany: TU Darmstadt -
> > http://www.kom.tu-darmstadt.de/en/research/research-areas/peer-to- 
> > peer-networking/overview/p2p-network-games/
> > ),
> > it seems that nobody was able to solve the problems.

> > May I can ask the GNUnet developers about their opinions and  
> > suggestions how
> > to reconcile Performance / Latency and Security at the same time ?


I would say that the problems are still being worked on. If you  
ordered the
requirements you listed above, security is probably the number one  
problem,
followed by latency. Allowing players to run the game as a  
distributed computation
allows them to cheat in many new and interesting ways. Without  
solving those security
problems, P2P MMOs will not be viable.

Latency is the second issue. For FPS games, you want an RTT of about  
100ms,
with an RPG you're looking at 250ms, and for RTS you're looking at  
around 500ms.
Once you start routing packets over a P2P network, you add multiple  
end-to-end
hops, which adds up quickly.

As far as commercially successful P2P games, the only documented one  
I know
about was Age of Empires, which Bettner and Terrano described in  
"1500 archers
on a 28.8: Network programming in the Age of Empires and beyond". Not  
sure if
they continued the P2P approach with successive games.

In my opinion, a possible way to reconcile latency and security is  
through the
hierarchical organization of peers in the network. This could be done  
by using
GNUnet to form a structured overlay for message dissemination. GNUnet  
already
adds some overhead for security reasons, but some of this is needed  
in any case
for protocol security in a P2P setting.

Chris