FF14 has a weird server architecture, at least by my understanding of how this should work. There are 3 data centers which impact overall latency, and each holds 8 worlds. To say that there’s been an uptick of players is an understatement… and the servers are struggling. Queues are present daily (you see your position, but not time estimate). S/E tried to fix this with a near 12 hour patch on Monday, with no real success (nor failure!)
As of this draft, of the 24 worlds, 12 of them will not allow new characters to be rolled – and one data center is completely locked. Side note, I transferred to this data center with less than 24 hours before they locked it…blame me I guess.
My field of work intersects with DC architecture. I know there are a lot of efforts to move out of the mid-00’s data center design, into the PaaS-type of design. This is not easy, as there are often specific lines of code in software that relies on specific OS/HW configurations. The ol’ hook/pull coding nightmare that all the cool kids thought was going to save the planet with their custom code. Upgrading is hard, you can’t break anything, and odds are you can’t fully test the load until you turn on the switch.
Yeah well, we’re still picking up the pieces of those decisions… and we’re going to be in a long tail of migration for a good chunk. Most large enterprises still have legacy code somewhere that they just can’t port over for A/B/C reasons after all. And in the gaming space, this particular code portability is the primary driver for sequels. No need to port if it’s brand new code!
Side note – a few years ago WoW updated their database architecture, which had a pretty darn big impact on skill calculations. They were using a DB structure that dated from Vanilla (well, more like 4 years prior to launch), and it was quite painful to move forward. They did it though! And the overall system load (if you read those reports) was significantly reduced. That requires some serious planning and smart people to deliver.
FF14, from the information that’s published, is talking about OS configuration issues and replacing some equipment with HPC. I do understand why they would not be moving to commercial cloud… that is a really bad idea for an MMO. Commercial (or even semi-private) cloud is absolutely not the silver bullet that some folks may believe – just like a buffet is not the best type of restaurant for all your food needs. Given the on-going supply chain issues, this is going to be a long-term effort to expand given capacity, let alone upgrade the components.
Hats off to S/E for the forthright nature in this issue. Fingers crossed that this is actually something they can actually address before November. Cause if queues exist before launch… well delaying launch is going to be right on the table.