You gross 200 Million in subscriptions a month why are your servers junk?

[spf1200]

All these countless threads whining about the servers only prove than most of the game's player base doesn't understand how servers work. More money doesn't mean they are able to get more servers in their data centers.

[Edax]

All these countless threads whining about the servers only prove than most of the game's player base doesn't understand how servers work. More money doesn't mean they are able to get more servers in their data centers.

/sarcasm It's almost as if there's a problem.

[Adrestia]

All these countless threads whining about the servers only prove than most of the game's player base doesn't understand how servers work. More money doesn't mean they are able to get more servers in their data centers.

Are you . . . quite certain you know how money works?

You're going to run out of cash before they run out of rack space.

[Catwho]

How the login queues work on a server with a stable population: You go to log in. You are number 29 in queue. You wait a minute. You are logged in.

How the login queues work on a server with a large population that is paranoid about login queues: You go to log in. You are 1,029 in queue. There are 6,000 people already on the server, of which 2000+ are using the crafting cheat to stay logged inn past the 30 minute auto disconnect. The other 4000 are playing. Occasionally one disconnects and a spot opens up. The queue crawls.

SE handled this at 4.0 launch by implementing a 24 hour mass logout because the same thing happened on Balmung server. Their servers get unstable beyond 5000-7000 active connections per world. Upgraded hardware won't fix it; it's a core problem at the database level, I suspect.

[Adrestia]

How the login queues work on a server with a stable population: You go to log in. You are number 29 in queue. You wait a minute. You are logged in.

How the login queues work on a server with a large population that is paranoid about login queues: You go to log in. You are 1,029 in queue. There are 6,000 people already on the server, of which 2000+ are using the crafting cheat to stay logged inn past the 30 minute auto disconnect. The other 4000 are playing. Occasionally one disconnects and a spot opens up. The queue crawls.

SE handled this at 4.0 launch by implementing a 24 hour mass logout because the same thing happened on Balmung server. Their servers get unstable beyond 5000-7000 active connections per world. Upgraded hardware won't fix it; it's a core problem at the database level, I suspect.

Please explain, in one sentence or less:
1) Why would it be database?
2) Why aren't databases horizontally scalable?
3) If database performance isn't contingent upon hardware, what makes it work? Is it magic?

[jazo]

hey dude that knows about servers, Im a software consultant and have worked on companies that use big data, remember that is not only how many data can pass on to "servers", the write/read data from and to the database also has its limits, thats why we dont have infinite inventory slots, each data transfer counts, no matter how fast the server transfer is, the data can only be handled so much. That is why they are opening new servers and data is balanced on them.

[Catwho]

Please explain, in one sentence or less:
1) Why would it be database?
2) Why aren't databases horizontally scalable?
3) If database performance isn't contingent upon hardware, what makes it work? Is it magic?

The following is speculation based on my limited technical understanding of SE's infrastructure.

1) They are using the old 1.0 RDBMS system instead of a proper modern NoSQL solution.
2) See point #1 - RDBMS that has been forced to emulated NoSQL by living in RAM in real time isn't as scalable as proper NoSQL solutions with sharding and such.
3) Hardware always has a point of diminishing returns, and they upgraded the servers twice before already - when the NA data center moved from Montreal, and again when Stormblood came out.

(I work in software development but I'm just the UX / business analyst. I do know that our old Oracle SQL database chokes out at around 10,000 active threads and we are nowhere near as intensive as an MMO.)

[Adrestia]

The following is speculation based on my limited technical understanding of SE's infrastructure.

1) They are using the old 1.0 RDBMS system instead of a proper modern NoSQL solution.
2) See point #1 - RDBMS that has been forced to emulated NoSQL by living in RAM in real time isn't as scalable as proper NoSQL solutions with sharding and such.
3) Hardware always has a point of diminishing returns, and they upgraded the servers twice before already - when the NA data center moved from Montreal, and again when Stormblood came out.

(I work in software development but I'm just the UX / business analyst. I do know that our old Oracle SQL database chokes out at around 10,000 active threads and we are nowhere near as intensive as an MMO.)

I'd probably argue that they're better off with SQL than NoSQL, given how they're likely to use their data. This isn't big data that needs to be passed through an aggregation pipeline. Relational tables should actually be more performant for what a game needs.

There's actually really good multi-master replication and performance clustering out there today. Not knowing whose DBMS they're using, I couldn't exactly specify a vendor/solution, but you might be interested in taking a look at something like Percona. Their clusters actually do feel kind of like magic. Now there are definite price concerns if they're using, say, Microsoft SQL Server. The MS approach is basically that if you want access to anything good, you pay through the nose, and it becomes hard to ever back down from it to the cheaper licenses. However, it's out there, and SE is a very profitable company. Businesses don't need customers defending their right to profits. They do that part, and we decide when it's time to provide more value for our dollar.

SQL is also incredibly CPU- and memory-bound, and I will purchase and eat a moogle hat if they're running the kind of hardware you bankroll when you actually care. That said, it's possible to write *very* bad SQL that will cause performance concerns to grow at a scale that no hardware can keep up with. Thankfully, that's code, and it can be rewritten. If I thought it would help (I don't; these are different skillsets), I'd sacrifice a dungeon for a backend performance increase. But the only thing that can and should be sacrificed is SE's operating income. This won't happen as long as their fans continue loudly defending the indefensible, however. They have no incentive to provide a better product if people are paying the same for this one.

You're not "just" a BA! We'd kill for another where I work right now. Or two. Personally I'd take 10.

[ChewieFlakes]

How many crappy expansions launches do they need to do before upgrading their garbage servers? Is this an indie company??

[Valkyrie_Lenneth]

How many crappy expansions launches do they need to do before upgrading their garbage servers? Is this an indie company??

What issues? No crashes beyond a single ddos attack on Friday. No login queues for us here, no real bugs that I've run into. All in all it's the smoothest launch yet, and probably the smoothest one I've ever seen.