ELI5: Why the absolute hell did SE not use Azure or AWS as a temporary solution?

[LianaThorne]

Isn't failing to do anything for the last nine months to improve servers to prepare for this launch when they have the data to show their MoM growth and historical data to show the increase in returning players for an expansion even worse? It's not like the idea is "do it in 2 days" but they had MONTHS to prepare for this launch and have utterly failed in that regard.

You do realize that this chip shortage has been around since at least the summer of 2020 yeah? And that the big boom in the playerbase didn't happen for months afterward? Did you expect them to be psychic and predict this happening back before the shortage began?

[Rhysati]

They still got to play and didn't get kicked after six+ hours in queue. That's still an improvement over this shit show.

You clearly haven't played it.

Destiny 2 runs on AWS and costs $0/mo.
For Honor runs on AWS and costs $0/mo.
The Division 2 runs on AWS and costs $0/mo.

None of those are MMOs. They all use lobby-systems and instancing.

[Merkhen]

getting booted and having multiple connection issues mid-play session is far worse, especially since you'd still be queueing to get back in, even if not for as long.

don't try to argue with people that say that queues are worse than any game-breaking bug, lag, or connection issue while playing(I have PTSD from Neverwinter and archeage), they're clearly delusional.

[Semirhage]

Moving something from a datacenter to a cloud provider isn't something you "just" do. There are tons of variables involved, and shifting projects between these two types of platforms is a full time job for full teams of people. Compounding this, FFXIV wasn't designed to run on a public cloud provider. Their backend code isn't organized in a way that AWS or Azure expect things to work. Services don't communicate with cloud servers in mind, etc. It's like going to Chicago and expecting a map of New York to tell you where everything is.

I do this for a living; moving entire projects into the cloud takes months to years for *well* architected projects. I have no doubt they looked at what it'd take to move XIV's ancient spaghetti code onto AWS and concluded it would be easier to just make a new MMO designed on scaling cloud servers from scratch.

[Larirawiel]

Destiny 2 runs on AWS and costs $0/mo.
For Honor runs on AWS and costs $0/mo.
The Division 2 runs on AWS and costs $0/mo.

WoW used AWS to relieve player congestion.... Did you do any research before posting this?

https://venturebeat.com/2018/03/19/a...ame-companies/

"to handle the ebbs and flows of demand on cloud-connected data centers"

It seems that the games do not run at 100% on AWS. Second, those games are not MMOs. And third, the server software must be developed in mind that it will propably run in cloud instances. If not then the server software will run very very VERY slow.

A cloud instance is more comparable to a Raspberry Pi than to a Xeon/Epyc-server. But you can rent thousands of those Raspberry Pi-like instances on demand and spawn them very quickly. And that is a problem. It is very unlikely that one cloud instance can handle more than a dozen players at once. And you do not want that only one dozen players are allowed per zone. Because it would not be an MMO anymore. And no, you cannot take 20 cloud instances and make one big instance of them because cloud instances are strictly separated from each other. You need to synchronize them permanently via network. And permanent synchronization creates a lot of network traffic.

Running on AWS may work for small group loot shooter but not for MMOs.

Cheers

[Zileth]

They still got to play and didn't get kicked after six+ hours in queue. That's still an improvement over this shit show.

However its lag lag lag lol I guess you like failing and dying due to lag lol

[Minsue]

Ask New World players how stable AWS is to actually play on.

Yeah it's not great.

[Akiudo]

Whats the point of your thread if you dont want people to click on them?

oh but he does want people to click on them. just not people that tell him he is to full of himself, only people that help him build the perfect circlejerk.

[Thoosa]

https://venturebeat.com/2018/03/19/a...ame-companies/

"to handle the ebbs and flows of demand on cloud-connected data centers"

It seems that the games do not run at 100% on AWS. Second, those games are not MMOs. And third, the server software must be developed in mind that it will propably run in cloud instances. If not then the server software will run very very VERY slow.

A cloud instance is more comparable to a Raspberry Pi than to a Xeon/Epyc-server. But you can rent thousands of those Raspberry Pi-like instances on demand and spawn them very quickly. And that is a problem. It is very unlikely that one cloud instance can handle more than a dozen players at once. And you do not want that only one dozen players are allowed per zone. Because it would not be an MMO anymore. And no, you cannot take 20 cloud instances and make one big instance of them because cloud instances are strictly separated from each other. You need to synchronize them permanently via network. And permanent synchronization creates a lot of network traffic.

Running on AWS may work for small group loot shooter but not for MMOs.

Cheers

Apparently it does: https://aws.amazon.com/blogs/gametec...rpg-migration/

[LittleImp]

AWS is horrible, and migrating their services to AWS on such short notice is extremely unrealistic. The amount of resources they would have had to commit to offset a few days of player discomfort were probably assessed by the business division as not being worth it, and would have likely facilitated additional delays.

Source: I am AWS certified.