4 ms·

Ok so then they appear to be "misusing" AWS. It's built for exactly the kind of problem they are having. Kudos for getting on it but they appear to have mis-des

Ok so then they appear to be "misusing" AWS. It's built for exactly the kind of problem they are having. Kudos for getting on it but they appear to have mis-designed their architecture then or something else. I'm not sure what's worse, as the author intimated, making a massive online game not using cloud tech, or using cloud tech and still suffering massive server resource shortage and being unable to scale anyways due to poor architecture?

nemothekid 14y ago
Using AWS/cloud service/buzzword-of-the-month doesn't magically make your scaling problems disappear.
- mindstab 14y ago
  Clearly as demonstrated by this mess. What is your point exactly? They are doing it wrong, however they are doing it. OP suggested they use AWS (which alone as stated doesn't solve the problem). Then first commenter said they are using AWS. I simply quiped ~ "more like misusing". Where in there is something wrong? Unless 30+ minute wait times is good design?
  - nemothekid 14y ago
    "Misusing" is wrong as well because no one is aware if they aren't doing things by the book. This isn't the first time a company has launched an MMO only to have it crash multiple times at launch. Starcraft had similar problems, and Blizzard STILL didn't get it right when it came time to launch Diablo. Trivializing Diablo's issues to "well Blizzard should have used AWS instead of managing their own servers" is just wrong, and doesn't get anyone anywhere. Its clear launching MMO's is hard. I'm not even sure the author is an authoritative figure on the subject - his service most likely grew slowly over months, rather than a million people at once. EA's issue could be completely unrelated to the fact they used AWS or not (or if they implemented DRM or not). Again, games crashing at launch isn't a new phenomenon. AWS allows you simply add and remove machines. If EA has 100 machines at this point, there is nothing AWS can do. It may even be possible their AWS infrastructure is on point and the problem relies elsewhere. In any case, we will never know if people just quip "well they should have used AWS/Postgres/MongoDB"
kevingadd 14y ago
Many problems are not magically solved by throwing more machines at them. The problem Minefold's blog post lazily describes as 'single point of failure' actually encapsulates certain classes of problems that are intrinsically singular; as it happens, friends lists are one of them. No matter how much horizontal and vertical scaling you do, it will eventually be possible for a player to create a friends list that has so many friends on it that operations on it end up becoming prohibitively expensive. Friends list operations end up effectively being JOINs.
- mindstab 14y ago
  So your theory is that problems with friends list are keeping people in 30+ minute queues to play the game? Because otherwise you are straw manning me. I agree, just throwing AWS at the problem doesn't solve it, as demonstrated by this catastrophe. On the other hand, many people, including OP have managed to build scalable gaming systems on top of AWS where EA/Maxis for the moment appear to have failed quite dramatically. Yes it's a hard problem, but it doesn't seem off base distilling the OP's argument down to "they have done the cloud wrong".
  - kevingadd 14y ago
    It's not a straw man, 'Better friends lists' is one of the sections in the original article.

nemothekid 14y ago

Using AWS/cloud service/buzzword-of-the-month doesn't magically make your scaling problems disappear.

mindstab 14y ago
Clearly as demonstrated by this mess. What is your point exactly? They are doing it wrong, however they are doing it. OP suggested they use AWS (which alone as stated doesn't solve the problem). Then first commenter said they are using AWS. I simply quiped ~ "more like misusing". Where in there is something wrong? Unless 30+ minute wait times is good design?
- nemothekid 14y ago
  "Misusing" is wrong as well because no one is aware if they aren't doing things by the book. This isn't the first time a company has launched an MMO only to have it crash multiple times at launch. Starcraft had similar problems, and Blizzard STILL didn't get it right when it came time to launch Diablo. Trivializing Diablo's issues to "well Blizzard should have used AWS instead of managing their own servers" is just wrong, and doesn't get anyone anywhere. Its clear launching MMO's is hard. I'm not even sure the author is an authoritative figure on the subject - his service most likely grew slowly over months, rather than a million people at once. EA's issue could be completely unrelated to the fact they used AWS or not (or if they implemented DRM or not). Again, games crashing at launch isn't a new phenomenon. AWS allows you simply add and remove machines. If EA has 100 machines at this point, there is nothing AWS can do. It may even be possible their AWS infrastructure is on point and the problem relies elsewhere. In any case, we will never know if people just quip "well they should have used AWS/Postgres/MongoDB"

kevingadd 14y ago

Many problems are not magically solved by throwing more machines at them. The problem Minefold's blog post lazily describes as 'single point of failure' actually encapsulates certain classes of problems that are intrinsically singular; as it happens, friends lists are one of them. No matter how much horizontal and vertical scaling you do, it will eventually be possible for a player to create a friends list that has so many friends on it that operations on it end up becoming prohibitively expensive. Friends list operations end up effectively being JOINs.

mindstab 14y ago
So your theory is that problems with friends list are keeping people in 30+ minute queues to play the game? Because otherwise you are straw manning me. I agree, just throwing AWS at the problem doesn't solve it, as demonstrated by this catastrophe. On the other hand, many people, including OP have managed to build scalable gaming systems on top of AWS where EA/Maxis for the moment appear to have failed quite dramatically. Yes it's a hard problem, but it doesn't seem off base distilling the OP's argument down to "they have done the cloud wrong".
- kevingadd 14y ago
  It's not a straw man, 'Better friends lists' is one of the sections in the original article.