Hacker News Reader

Comment by WClayFerguson | Hacker News Reader

I've been telling developers for years if you build your infrastructure on top of tons of 3rd party services like Google Cloud or AWS then you basically don't own it, and can be shut down any day, against your will.

Web companies should build on "Linux" only, as their base platform, so they own everything above that. I was a fan of Parler but building on AWS was a huge mistake, and they should have realized they were "owned" from day one.

And to any other CEO/CTO making the same mistake: "Beware, you do not have control over your company."

View on HN

It isn't just getting kicked off the platform that you should worry about. You have zero control of long term pricing. Developers sometimes even develop without doing any math on what the services they are using cost in year one. Ask people how often they have already had to go back and rework code and shutdown servers because the bills coming in were out of control. The myth that you can just migrate if costs go crazy is completely nuts.

Initially when I saw the EC2 micro prices I was like "wow that's a fraction of a cent" but then you multiply it 750 times in a month like oh...

It is increasingly, vanishingly hard to not use the big three cloud services. Mainly because any platform or tool you might use is _also_ hosted on these clouds.

You can build on Java + SpringBoot + MySQL + Linux + Docker Compose, then what else do you really need?

I mean there's tons of technology stacks to choose from, but people get lazy and take the easy way out rather than learning to scale servers themselves.

Get on Linode. It's base on Linux instances. If you build on Linux you are "free" and not "owned"

... what else do you really need?

A proper database.

There's a lot of DBs to choose from: MongoDB, MySQL, PostgreSQL. What you want to avoid is anything from Amazon or Google ANY other similar service.

Is that supposed to be a dig against MySQL, or did you just miss that in the list?

Looks like you misspelled "NoSQL" :)

Linode and Digital Ocean docs also are leagues ahead of AWS. They cover a lot of stuff. I didn’t even know what fail2ban was when one of my VPS’s got hacked. Went through all their docs, they cover a lot and really helped me understand how much I didn’t know while at the same time providing guides on getting up to speed.

AWS won’t tell you about those things because they sell you security, so you remain a dumbass about those things indefinitely.

Exactly right. lol.

If you are doing ssh config properly, then you don’t need fail2ban or its ilk.

Fail2ban protects more services running on a server than just SSH.

Most everything I develop is in .NET and will run comfortably in Nginx, IIS, or an Azure App Service Resource.

It's not that hard.

In my experience most developers just inherit the infrastructure and even the base project from someone else. They've never stood up a server or configure anything from scratch so they have no idea what they're doing.

I've been telling people for ages that, even if they want to use a pre-built solution, they should at least set up the base systems a few times to get a better felling for them and.

that logic is a vicious cycle

Why can’t they just backup regularly and redeploy somewhere? Isn’t it as easy as backing up a DB and restoring?

When you build on top of "linux", yes. However, many cloud services are built on top of S3 or App Engine or other proprietary services.

This means that they can only be run on that provider.

They can be faster to develop, at the risk of not being able to move later.

S3 is not that hard to replace. There are plenty of services like backblaze which provide S3 Api. You can spin up your own service with something like minio.

Getting S3 reliability or availablity is lot harder ofcourse, but most apps don't need that.

If you restrict yourself to only the set of AWS functionality that is common to all hosting providers, you can do that.

But not only is that very hard to do, it basically negates the premise of AWS in general. There are also lots of areas that you can’t do this, such as IAM and network rules, so you have to build an abstraction on top of it that’ll work with open source tools on a separate host.

Premise of AWS for me: reasonably priced (not cheap!) linux VM, close to the backbone, essentially total control of the VM, easy provisioning of new disk storage when required (i.e. EC2 with a smidgeon of EBS)

The rest of it? I think I understand why some systems need/want it, but no thanks.

Depends on what you’re doing.

At the scale $EMPLOYER works at, AWS built in services are key. Things like DynamoDB global tables are extremely hard to run on your own, and things similar to AuroraDB and Kinesis can be run on your own but they take up a lot of personnel hours better spent focusing on product. We could probably eliminate those services and self-run, but it would be an extremely unwise usage of company time and money.

But we also run services in multiple continents with the need to handle clients doing silly things like getting into airplanes and flying into another continent. Your situation might not benefit as much from AWS specific services, so YMMV.

> We could probably eliminate those services and self-run, but it would be an extremely unwise usage of company time and money.

Unless $EMPLOYER was Parler ...

More precisely, we could do it, but it would probably take years and the result would be inferior to what we have today.

They don't even need their own backups anymore.

Linux doesn't help here. You can have your own servers, but then ISP can shut you down and Mastercard and Visa will ban you from the online economy at will ...

If Silk Road (and its spawn) manage to stay online until the government shuts them down then these more-legal companies can surely find a way.

Sure if your users can use crypto and tor you can stay open for a few years... but that isn't the same problem.

Linux is much better than relying on proprietary services. Nobody said Linux solves all the other problems in the world. lol.

I don't know what do you mean really, what do you think Amazon uses for their clouds, Windows 95?

according to parler they did what you suggest

Parler used AWS. I suggested avoid AWS. What am I missing. I might be missing something.

AWS is a suite of services. One of those is EC2, which is just virtual servers hosted by Amazon/AWS. A virtual server is no different than a physical server from the perspective of the guest operating system, regardless of if it is Linux or not. Linux is just the software running on the server, it has nothing to do with ownership. The only way "running on AWS" is problematic is if you tied yourself to their non-generic services like ECS or Beanstalk.

tl/dr: building on AWS and building on Linux are not mutually exclusive. One represents hardware the other software.

We used AWS and EC2 at my last job so I've done about 4 years worth of what you're describing. Definitely if you use Amazon as your way of getting a Linux instance that's fine, because then you can just move to Linode or any other hosting service because Linux is the one thing they all support.

When I say don't use "AWS" I should've been more clear I'm talking about their proprietary stuff. Basically you should be able to run on a plain Linux as your base.

Parler didn’t use any proprietary AWS services.

The Parler CEO initially said they would be offline for a whole week, in order to “rebuild from scratch”.

Reply on news.ycombinator.com