Quote:And lastly, nobody, and I mean nobody cares about network code. Players definitely don't. The only thing that matters when developing an MMO is content. But if it then runs on a lean mean single machine or a bunch of "poorly" optimized mediocre machine is something nobody will ever know or care about.

I think this is were we if not disagree misunderstand each other. To start with I care, and I care a lot. Now, I'm a rare breed of course and I guess you mean the participant not the developer. But I cant stop to wonder that developing anything less than superior or excellent is wrong by nature? I mean, I do understand the time and effort constraint, not to mention maintainability issues, but I've always settled for nothing less than the best.


Now I don't want to head into a discussion about what is "the best", but I wouldn't really want to take part in a solution were you have some mediocre middle ground, and basically solved "issues" by buying even more hardware. However, the idea of a network based on load balancing to handle high pressure, and were no single machine is irreplaceable sounds kind of ideal to me.

So to summarize things one could say that no MMO implementation is based around the cleverness of a single application, it is instead the joint effort (or if you wish "harmony") of networked computers (a cluster) which all work as a unit. Were the greatest of challenge is to come up with a method for distributed computing of the massive persistent state.

I think that came out rather nicely, eh?

Also you make an excellent point in that, hardware can be cheap in comparison to development. And I did read a while back about this company which entire infrastructure was based around an old replicable and redundant Linux cluster. Were their economic system was so crucial to their business, if it were to go down, so would probably the company as well. And their solution was to go for many, rather than a single "silver bullet" (Since there is no such thing as a silver bullet...). And it turned out to be cheaper to just replace a malfunctioning machine than start a debugging session.

Quote:But I cant stop to wonder that developing anything less than superior or excellent is wrong by nature? I mean, I do understand the time and effort constraint, not to mention maintainability issues, but I've always settled for nothing less than the best.

In that case you've never worked on a commercial project with a budget and a deadline involving hundreds of people.

Quote:Now I don't want to head into a discussion about what is "the best", but I wouldn't really want to take part in a solution were you have some mediocre middle ground, and basically solved "issues" by buying even more hardware. However, the idea of a network based on load balancing to handle high pressure, and were no single machine is irreplaceable sounds kind of ideal to me.

google.com. It costs billions of dollars to build such a system, but they have it. Much of it is also documented.

They use pretty mediocre hardware, and pretty mediocre software. Much of it is perl, python, java, all of them very "slow" and "inefficient" software. If they run out of space, they simply put another 5000 computers on the grid. And if 100 of their cluster nodes die a day, it's an operational write-off.

They do however develop some pretty low level hacks, but not for performance reasons, but to get around limitations of mediocre operating systems, which were never used for such huge sizes.

Quote:So to summarize things one could say that no MMO implementation is based around the cleverness of a single application, it is instead the joint effort (or if you wish "harmony") of networked computers (a cluster) which all work as a unit. Were the greatest of challenge is to come up with a method for distributed computing of the massive persistent state.

If you come up with groundbreaking solution to this, you can sell it to NSA for millions, even billions.

Such tools would be invaluable in physics, cryptography, military simulations, anywhere really. And those areas have budgets of tens of billions a year.

Distributed computing is an old topic. And today, a very apealing one. Which also implies there's a lot of knowledge out there already.

But it has nothing to do with sockets or optimized code.

Quote:In that case you've never worked on a commercial project with a budget and a deadline involving hundreds of people.

Actually I have (not hundreds of people though more like a few), but I never really liked this middle ground thing. But you do what you can with the time you have, right? ;)

Quote:Distributed computing is an old topic. And today, a very apealing one. Which also implies there's a lot of knowledge out there already.

But it has nothing to do with sockets or optimized code.

I guess pretty much started out in the completely wrong end! But this thread has been very enlightning, and I thank you Antheus for a very rewarding discussion.

Best Regards,
John