What EA could have done to help Sim City | Hacker News Reader

What EA could have done to help Sim City

blog.minefold.com

33 pointschrislloyd19 comments

Open articleView on HN

I love armchairing as much as the next guy, but this is lazy spam, plain and simple:

nslookup api.simcity.com

nslookup socket.simcity.com

nslookup update.prod.simcity.com

Right now these are returning CNAMEs to amazon web services and akamai for me, and I'm told they were days ago as well.

If you're going to talk about how EA should have set up SimCity, maybe you should at least look and see how they ACTUALLY set up SimCity and make sure they didn't do the things that you claim would have fixed their problems?

EDIT: Also, getting a launch right for a tremendously mass-market product is hard work. Until you've actually done it, it's kind of shitty to stand back and shout simple advice like 'use the cloud' or 'use modern devops practices'. I think that's what really bugs me about this post: The minefold guys probably do have useful bits of scaling advice to share here, but they're burying it under lazy swipes at a big competitor that really aren't justified.

Exactly. We can get into a discussion of AWS best practices but without more detail on their application's structure their usage of AWS right out of the gate shows that somebody at EA/Maxis has their head in the right place.

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?

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

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?

"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"

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.

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".

It's not a straw man, 'Better friends lists' is one of the sections in the original article.

a) Thanks for the legwork. Its easy to point fingers.

b) Singling out EA for bad tech is like beating a dinosaur... it's been dead for millions of years damnit! EA has not been know for making quality, reliable, stable games. In fact its exactly the reason I wouldn't buy anything they make. And usually won't even pirate it because its not worth the headaches of crashes and friends.

How about: 1. Not require always-on DRM for an almost completely single player game.

Minecraft seems to strike the right balance here. Authenticate players on public servers to avoid spoofing other accounts, but also allow people to run their own private servers and disable online mode. Granted this enables piracy but I'm sure it hasn't put a dent in their sales.

Even if it has put a dent in their sales, Notch doesn't seem to mind.

As much as I agree with the sentiment, it hasn't actually affected their sales. If consumers actually cared about it then there would have been far fewer sales and the servers wouldn't be overloaded in the first place!

No shit. And let's quit going along with the fiction that the online requirement was for any other purpose.

"If the original title begins with a number or number + gratuitous adjective, we'd appreciate it if you'd crop it. E.g. translate "10 Ways To Do X" to "How To Do X," and "14 Amazing Ys" to "Ys." Exception: when the number is meaningful, e.g. "The 5 Platonic Solids."

Otherwise please use the original title, unless it is misleading or linkbait."

Via http://ycombinator.com/newsguidelines.html

Thanks, fixed.

Yeah! Why didn't they elastically scale with the cloud? #fail

They don't need help, they already got your money.

Reply on news.ycombinator.com