Sun recently announced
href="https://games.dev.java.net/">Project Darkstar, a variant of the Sun Grid.
According to the project website:
"Project Darkstar", is a research effort aimed at simplifying the process of developing multi-player online games
that can be deployed on a massive scale and made available to players using virtually any client device.
The first technology release under this project, the Sun(tm) Game Server, is a game agnostic, platform agnostic
server technology that provides online game developers the ability to create a wide variety of games that can be
provisioned through a single server platform.
One of the problems that game developers have had in the past with making MMOGs has been in relation to network
architecture -- every game company needs to run their own servers, deal with the issues of scalability and so forth.
While I occassionally have my doubts about the network programming that is found within games, for the most part,
it's fairly competent and solid stuff -- the problems that I've experienced when it comes with network scalability
has more to do with the financial side of the business making engineering spec -- such as finance managers deciding
that due to cost engineers need to cram 2000 users on a single server rather than the 500 user max the engineers
determined.
I trust engineers more than financial managers when it comes to scaling network architecture, and I believe it to be
to the benefit of game companies to have network programmers to keep unrealistic server expectations in check. While
I have no doubt that Project Darkstar will be fabulous for Java-based casual games, I have my doubts as to whether
it will be a viable platform for desktop based games.
Okay, let me clear up a few of your misconceptions
Project Darkstar is NOT part of the Sun Grid effort,. While we cna run on the SUn Grid, we also run on any other reasonable backend.
Secondly, Project Darkstar is a SUn Labs project, We are all engineers. In fact, as the Architect of Project Darkstar I came from the game industry, My code is in games such as GEX, TITAN, The Horde and Blazing Dragons from Crystal Dynamics. I was the Senior Game Integration Engineer at the Total Entetertainment Network and wrote internet code for games such as DukeNukem3D, Total Anihilation and Dark Sun Online.
You are certainly free to have your doubts, But they were not shared by the industry experts who viewed our simultaenous GDC demos of a causal game, a fast action physics based racing game and an MMO demo all running simultaneously on the same system running the Darkstar-SGS SDK EA1
At Sun, we built the largest, most reliable on-line games in the world. They are called Stock Exchanges. Project Darkstar is bringing that expertise to solve the game industries online problems.
Jeff,
I think you misunderstood me.
I have a certain amount of faith in your product solution, I don't have faith in managers who arbitrarily dictate spec, which is what is likely responsible for most game companies' failings when it comes to the network architecture for massively multiplayer games.
I think it's great that Project Darkstar is out there and available. I'm just pointing out to people that the greatest obstacle that game developers face in developing a MMOG is the cost of the hardware -- the network code that is being run on the servers is cheap comparatively speaking.
The game industry's online problems are not scalability issues. The issues and problems experienced by gamers and developers arise more from bandwidth and their graphics hardware.
Hi Mike!
You are absolutely right that hardware costs are a big killer in online agmes (although dont underestimate the time and difficulty of it building the server side sofwtare either.)
Project Darkstar addresses this in two ways:
(1) The current Shard and Region systems for MMOs are MASSIVELY inefficient. When a region is not fully occupied, you waste cycles. When a region is totally empty you waste machines. This cost doubles if you want any degree of reliability as you need to run mirror servers to take over when main serves fail.
Project Darkstar eliminiates that waste. We ALWAYS load balance across ALL resources. This is a result of our having eliminated shards and regions as load-balancoing constructs. Instead, machiens are assigned to users and the data coems to them as needed. We need no backup mirrors because EVERY Dakrstar "slice" is a backup for every other slice already.
(2) The huge cost of these games is also that every game ash to build its own machine room. We eliminate that. As a container system, Darkstar actually load balances acorss multiple games installed into the same back end simultaneously.
This means a big publisher can run a single machine room for ALL their games. A small publisher can go to a Darkstar hosting provider and outsource it all-- the same way J2EE develoerps go to J2EE hosting providers today.
So you are right that machine costs ARE an issue in the MMO space today (thoug I dispute that theya re the only one)-- and Darkstar is specifically designed to address that cost of equiptemnt along with the other major MMO issues!
I hope that clears up any remaining misconceptions.
Regards
Jeff Kesselman
Darkstar Architect
--------