[vworld-tech] Combining networking modes?

[Martin Piper]

Oops. Well that'll teach me to double check the sender field. I'll make the
"didn't have my wakeup cup of tea" excuse, sorry admins and everyone. ;)

While some of the various middleware choices for using C++ classes to try to
solve network data organisation are similar, they propagate C++ class member
data, there are differences in implementation. For example ReplicaNet
encourages programmatically defining class member data/functions to
propagate, as well as using a the ROL script language. ReplicaNet also
allows raw data packets to be sent for when the user really has to use such
things themselves, for chat server messages for example.

I'll refrain from going in to too much depth however, since it might be seen
as being a bit off topic. :) (I can talk for hours on this topic.)

In relation to your main question, it is an interesting problem. If each
battleship was fully populated with players then that is 80,000 players in
the same world. To supply FPS experiences to that many players, who could
interact with each other, would require the use of quite a few server nodes.
This distribution of resources is required to help avoid the situation where
servers become slow to process events or congested with network traffic.

Players that are "close" to other players would tend to be grouped by the
network logic to be on the same server nodes. ReplicaObjects can be migrated
between nodes with very little disruption seen by the client. In tests where
player objects are voluntarily migrated between server nodes, during load
balancing, the client might see latency artefacts similar to dropping one
packet during a normal FPS game. This is hardly earth shattering since
packets are dropped during FPS games without people noticing much.

This is potentially good news for you because being able to group players
with close spatial relationships on the same server definitely helps reduce
lag. :) Most of the middleware solutions allow objects to be migrated in
this fashion and some even allow program control of this migration through
their APIs.

Depending on the hardware though an estimate on the number of players it can
update to give FPS performance needs to be made. Lets say a good spec server
could cope with 512 players and still maintain reasonable response and
network bandwidth. This would mean 157 servers sitting in a room. :) Not
impossible figures.

I hope this helps?


You might be able to tell from this that I love the concept of using
peer-to-peer networks and class based object replication and propagation. So
I might be a bit biased. :)


-----Original Message-----
From: Crosbie Fitch [mailto:crosbie at cyberspaceengineers.org]
Sent: Friday, 04 June, 2004 2:07 PM
To: 'vworld-tech'
Subject: RE: [vworld-tech] Combining networking modes?


> From: Martin Piper
> I'm sending the privately because it might seem like 
> advertising, but I know
> ReplicaNet (www.replicanet.com) would cope with that kind of game. :)

Seems a bit like Quazal's NetZ.

Any major differences?

Anyway, I'm primarily interested to know if anyone's done/doing a brute
force hybridisation, e.g. client/server MMOG supporting say 5,000
battleships with a local FPS system supporting up to 16 player crews.

So I'm not necessarily wondering what technology to use because I need to
develop such a game, I just use such a game as a possible example for my
primary interest, which is to know whether anyone has created such a
hybridised system.