Expanding Server Capacity Clarification

[Alhanelem]

But for corporations, that physical limit is much higher than what SE is using right now.

You don't really know how much bandwidth is available at their datacenters, nor do you know what kind of CPUs they are using, how much CPU usage their processes require, and how much of that one person's connection accounts for.

[Hulan]

You don't really know how much bandwidth is available at their datacenters, nor do you know what kind of CPUs they are using, how much CPU usage their processes require, and how much of that one person's connection accounts for.

Of course, but since we're discussing optimization, I am working under the assumption that they are doing "the best that can be done". I know how many packets my game is sending (it's printed in game after all), so I can formulate a rough estimate of the bandwidth utilized by ~5,000 users. Worst case scenario, they are pumping it all through a single connection, still puts it somewhere around 256Mb/s (~2Kb/s up/down * 5,000 * 25 servers). Assuming (again, I know, damn all of these assumptions), they're using OC-12 lines, which is the standard for mid-range corporate usage, they have 2488.32 Mb/s available. Now, generally speaking, those servers probably handle other, not-obvious traffic, but I'd say that is not a terrible estimate of their current position, bitrate wise.

If we're talking about CPU usage, that's just a matter of load balancing and distributing subprocesses to sub-servers in the cluster. Such a thing is not trivial, but it is also not impossible. It would take a few months, but it's not a solid wall, just a steep incline.

Now, please remember, I'm not saying that taking advantage of that extra space is easy, only that it exists. In fact, I originally made the exact opposite point. Adding more servers is the more efficient solution to the current problem. Adding more concurrent users is prohibitively difficult in the short term.