Linked by Thom Holwerda on Wed 30th Jan 2013 00:31 UTC
Games "EVE Online's complicated inter-corporate politics are often held together by fragile diplomatic treaties and economic agreements. So fragile, in fact, that a single misclick can lead to a fracas that quickly snowballs into all-out warfare. That's what happened to two of the spacefaring sandbox MMO's largest player alliances in the Battle of Asakai, a massive fleet vs. fleet onslaught involving 3,000 players piloting ships ranging from small interceptors to gargantuan capital ships. Straight from the wreckage-strewn outcome of the battle, we're breaking down the basics of what happened for everyone to truly fathom one of the biggest engagements in the game's history." The costs of this battle in in-game currency is, so far, 700 billion. While MMO's don't float my boat, I have to say that this is still pretty awesome. Penny Arcade looks at the technical details server-side, and what a battle like this does to the game's backend infrastructure.
Thread beginning with comment 550984
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[2]: Server Switching
by Brendan on Thu 31st Jan 2013 04:29 UTC in reply to "RE: Server Switching"
Brendan
Member since:
2005-11-16

Hi,

"As far as I can tell, EVE Online uses separate servers (where the client disconnects from the previous node and connects to the new node when the player "warps" from one solar system to another), and they're lying about it being a distributed system.


You just described a textbook example of a distributed system, but then criticized it for calling itself one... Why on earth did anyone mod you up?
"

If you want to stretch the definition of "distributed system" to include the "least distributed possible" cases; then you could pretend almost anything is a distributed system (all the way back to the old "telnet into a server" MUDs MUSHes and MOOs) and it becomes meaningless joke. I'm fine with that if that's what you want - let's call it a "barely distributed system" (no fault tolerance, no load balancing, etc; where the entire pile of crud has to be taken down for a few hours every week because they're too stupid to figure out live migration or even handle the complexities of symbolic links).

So; are they lying about it being a "more distributed than everything else" system?

- Brendan

Reply Parent Score: 3

RE[3]: Server Switching
by galvanash on Thu 31st Jan 2013 04:47 in reply to "RE[2]: Server Switching"
galvanash Member since:
2006-01-25

If you want to stretch the definition of "distributed system" to include the "least distributed possible" cases; then you could pretend almost anything is a distributed system (all the way back to the old "telnet into a server" MUDs MUSHes and MOOs) and it becomes meaningless joke.


Traditional MUDs... no. But there was some effort to make distributed system of MUDs

http://en.wikipedia.org/wiki/InterMUD

Note there is no mention of fault tolerence, load balancing, distributed processing, or anything of the sort.

I'm fine with that if that's what you want - let's call it a "barely distributed system" (no fault tolerance, no load balancing, etc; where the entire pile of crud has to be taken down for a few hours every week because they're too stupid to figure out live migration or even handle the complexities of symbolic links).


I agree with your gripe - only pointing out that fault tolerance and load balancing are only tangentially related to distributed systems architecture. Many distributed systems have none of those attributes, and many systems having those attributes are not distributed systems.

So; are they lying about it being a "more distributed than everything else" system?


If they run a large number of primarily independent servers that do most of their work on local independent datasets, only communicating to each other over narrow channels, then they are textbook distributed systems. The more state they share the less "distributed" they are...

Reply Parent Score: 3

RE[4]: Server Switching
by Brendan on Thu 31st Jan 2013 06:09 in reply to "RE[3]: Server Switching"
Brendan Member since:
2005-11-16

Hi,

"If you want to stretch the definition of "distributed system" to include the "least distributed possible" cases; then you could pretend almost anything is a distributed system (all the way back to the old "telnet into a server" MUDs MUSHes and MOOs) and it becomes meaningless joke.


Traditional MUDs... no. But there was some effort to make distributed system of MUDs.
"

I was looking at wikipedia's list of "distributed computing architectures":

http://en.wikipedia.org/wiki/Distributed_computing#Architectures

You'll see that boring old "client-server" (potentially including one client on one computer talking to one server on a different computer) is the first architecture on their list.

In my opinion, boring old "client-server" (including multiple clients talking to one server, and multiple clients talking to multiple separate servers) is just client-server and doesn't really qualify as a true distributed system.

Now; EVE Online (as I imagine it) is a slightly more complex case of client-server. I'd imagine that each individual client is talking to at least 3 different servers (one for chat, one for the economy/trade, and another for "objects in space"); but despite this it's still all just client-server, and still doesn't really qualify as a true distributed system in my opinion.

If they run a large number of primarily independent servers that do most of their work on local independent datasets, only communicating to each other over narrow channels, then they are textbook distributed systems. The more state they share the less "distributed" they are...


These are all practical considerations only (e.g. shared state and/or heavy communication tends to kill scalability/performance; and local independent datasets is the result of minimising shared state and communication).

What does qualify as "true distributed" is when multiple computers work together, rather then independently. Google (many computers working in parallel for each query), Wikipedia (front-ends, caches, databases, media servers, etc for each page request), BitTorrent, SETI@home, supercomputers.

- Brendan

Reply Parent Score: 2