Comment by nzoschke | Hacker News Reader

Comment by nzoschke | Hacker News Reader

Parent Full thread

The article is also talking about the software tools that glue everything together.

> Open source infrastructure refers to all the tools that help developers build software. On a deep level, it includes physical things like servers, but closer to the surface, it also includes things like programming languages, frameworks, and libraries.

I also don't follow the AWS -> not open logic.

I lease resources in someone else's data center. I use proprietary software services when they make building my business software easier.

AWS is not an turtles all the way down open stack, but their docs are good, have open source clients for all their APIs, and offer up standard Linux and Docker systems.

Convox is open source software. It makes setting everything up easy and ensures total access and visibility into your resources. It also adds open protocols like we hooks and syslog to integrate with your other systems.

You can't get the source to AWS to run it yourself, and you have to accept Amazons pricing, but it's still a very open ecosystem.

Providing open-source clients does not make AWS an open ecosystem. If anything, my sense is that the overall direction of the AWS platform has been to get more and more proprietary and closed with initiatives like Lambda and Redshift that can't exist at all outside AWS. Rather than being some indication of the openness of the platform, I'd say that having open-source clients and sdks simply makes good business sense.

Once you're up and running on the AWS platform with any significant infrastructure setup, you're basically locked in for life. I hope there are counter-examples, but I'm unaware of any significant migrations off of AWS in the last few years. Some of that is due to their overall pretty high service level, but it's also incredibly hard to replicate the AWS infrastructure on your own because the tooling either doesn't exist or is immature.

I think there's a really under-served niche of startups with AWS spends of greater than $200K/year that would love to move some or most of their infra off of AWS to achieve greater cost saving or unlock different product possibilities that can come with running in a datacenter or across different cloud providers. And would be willing to pay for services and tools to fill that need. I'm finding myself in that situation now, and even though we've been careful to not get tied in to most of the managed AWS services, we're still facing lock-in due to heavy integration with cloudformation. Without a good open-source or managed service alternative to just cloudformation, we're basically faced with needing to completely re-architect our systems if we move off of AWS. And full re-architecture initiatives tend to stay near the bottom of the Devops backlog if things are working.

Redshift is Postgres 8 something. It has proprietary extensions but I've had a really easy time getting data in and out in my limited usage.

I saw a local Lambda emulator today on HN.

But I do understand your concern. Software on AWS has an serious gravity and is hard to move.

I totally understand the CF challenge. I'm locked into that tool too. Terraform is great but doesn't really offer anything to replace the deepest parts of CF.

Which brings us back to the article. More funding for tools in the space would be huge for all of us.

Reply on news.ycombinator.com