Hacker News Reader

Comment by crazygringo | Hacker News Reader

In a business sense, absolutely not. Sure they could build it technically, but they'd lose money.

The issue isn't expanding data centers -- that's "easy". The issue is all of the software tooling, all of the gigantically massive platforms and services they have to not only build, but catch up to AWS/Google/Azure, all of which are moving targets.

AWS launched in 2002. Google Cloud (GCE) basically started in 2012 and only recently has finally mostly caught up to become a fully-fledged viable competitor, which took an entire decade. Meta would be too late to the party.

Sure they could launch something much smaller like a Digital Ocean competitor -- not a fully-featured cloud, mostly just servers and storage. But why would they? The margins are low and there's no strategic reason to do so.

View on HN

On one hand, I agree with you catching in all aspects would be near impossible.

However similar to Apple, a more focused approach could be viable and profitable much more quickly.

AWS has been around for a long time, but they also have tons and tons of feature bloat, and full of redundant services with different pricing. Should I use an ELB or an ALB? Fargate, ECS, EKS, code deploy? The list of services is overwhelming and confusing to most.

A new major player with a cleanly architected design from the start with a clear path to happiness could be successful. Though Meta is full of feature bloat so I don't know if they would be the right choice to execute that vision.

Your examples aren’t really that good. An ALB is a type of ELB (load balancer). AWS has multiple types, ALBs run on layer 7 and NLBs run on layer 4 of the network stack. They also have classic which are considered deprecated and NAT Gateways. None of this is redundant. Similarly Fargate is a type of long running task you could start and run on an ECS cluster. Think web service running in a container instead of a traditional group of EC2 virtual machines. Tartare tasks are also a higher level virtualization abstraction compared to using EC2 in your ECS cluster.

On the surface they all seem redundant because of the sheer breadth but when you start poking at them for a new solution that same breadth affords massive flexibility.

I think you are just helping make his point.

You can get flexibility or simplicity, pick your poison. Not all solutions have the same needs so the idea that there exists a single, simple out of the box solution for every problem is naive at best.

Yes and they were advocating for simplicity as a possible (starting) way for Meta.

Except the complexity wasn't created for the sake of creating complexity - it exists for a reason. You can go the simple route if you happen to be brand new to virtual infrastructure management. The problem is you're going to quickly outgrow your solution and then what? Besides, if you want simple you can go with Wix, Wordpress, or Digital Ocean.

I'm sure that at least some of the complexity in AWS exists because they can't go back and refactor their offerings when so many users depend on it.

“A monad is just a monoid in the category of endofunctors, what's the problem?”

There may be "niche" areas that are still big enough to be worthwhile where AWS is not very good and Facebook could compete. The obvious example to me is Machine Learning workloads. Facebook has lots of experience here, and AWS overall isn't very good at it. There are lots of smaller niche players with better offerings, and room for a big competitor. There may also be examples in other areas.

Also, AWS is very complicated now, so any new entrant that can build a simpler version from the ground up will be at an advantage

Launching a simple storage and services product would be the opposite of productive. They would be wise to lean into their strengths and offer things others arent.

I've written about it before, but facebook should offer its components as a service to startup social products. https://telegra.ph/Facebook-Social-Services-FbSS-a-missed-op...

Never, ever host your service with a potential competitor.

Like Netflix hosted on AWS?

Who says they are even a "competitor"? This like anti-evil, spam, localization, and login services can apply to a ton of products that are not direct competitors to FB. Sure FB can and will clone just about anything, but you can say the same thing about Amazon and Google and Microsoft, at which point you shouldn't use any cloud services.

> at which point you shouldn't use any cloud services

You may be on to something there.

Cloud services exist first and foremost to further skew the playingfield in the direction of the larger parties because the economies of scale they create will allow them to compete with you in legal ways that you will find hard to match. That is the reason why these large companies are willing to rent out their data centers to you: you are essentially allowing them to sell you value added bits whilst reducing their own marginal costs in the process. Double win for them.

The fact that they could do even worse doesn't matter much absent proof. But for Facebook I'll make an exception: that company is rotten to the core in a way that not much else is and I highly doubt that there is a barrier low enough that they wouldn't crawl under it.

They had Parse (which they acquired) and shut down. It wouldn't compete with AWS, but could have competed with Firebase if they had kept it around.

I would argue that google didn't catch up. Their product offering is vastly smaller and is the better for it.

Reply on news.ycombinator.com