nel-all
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Nel] Re:More news from the nevrax teams...


From: Daniel Miller
Subject: [Nel] Re:More news from the nevrax teams...
Date: Thu, 21 Jun 2001 18:47:37 +0200

To extend the explanation a little - the way we see things is this:

- The server sets for MMORPGs and the like have to be secure - that means
that any old Tom, Dick or Harry can't send a message directly to the economy
service to tell it that they've just won the lottery :)

- This means that the clients only have access to the front end services on
the front end servers and that they don't have access to the naming service.
(in order to connect to the front end services clients have to use the login
service as a go-between... there are examples in our samples directory
showing how this works)

- So - the naming service runs on a smallish LAN that connects the game
servers.  It broadcasts messages like 'the xxx service has just arrived at
address yyy' or 'the www service at address zzz has gone down/ disappeared'.
These are cached by each service to keep an up to date address book of the
other services that they can talk to.

- In the normal way we don't see services being started or stopped as a very
high frequency event.

I hope this helps.

Daniel.


--------------------------------------------------------------------
Hello,


> I have a question about the caching for the naming service. You say it
> automatically corrects caches when services move or are deleted. This
> doesn't sound very scalable, it every client querying the name server
> has to be contacted when something they asked about changed (especially
> over high latency links).


Right, but in fact, we developed the system to fit our requirement for our
game and in our system (in the shard), the naming service and all  ther
services are on a LAN, so we don't have high latency links.

regards,

Vianney Lecroart



reply via email to

[Prev in Thread] Current Thread [Next in Thread]