Ask HN: Is a “no-ops” architecture viable?
Like many developers I know, I am somewhat set in my ways when building new web projects - it's usually a bunkhouse nginx/express/postgres on an EC2, or Heroku if I am willing to spend the $$. However, with all the apparent advances in serverless technology and the mature DBaaS providers out there, it got me wondering if startups or other developers are realizing the promised benefits of these architectures? To me, it seems like Heroku is the easiest minimal-ops solution out there, but it can get pricey. Even a simple EC2 setup could require fiddling in the AWS console (or CloudFormation templates), ports & security settings, VPCs, DB backups, etc. In both cases you still would need to pay attention to resource allocation.
I see a lot of news about serverless architectures & frameworks, and other hosted services like auth, DB's, etc. Are any startups experimenting with building fully "no-ops" architectures (I use the word as a classification, I realize true no-ops is impossible)? Or at a certain point is the benefit of minimal ops outweighed by the overhead of needing to glue multiple separate services together? Heroku keeps adding features, allows more control, supports bigger companies and have many many plugins (elements I think they call it) you can install. There's already new PaaS with less features (and cheaper). You push something to a repository and they run it. I saw https://www.netlify.com/ mentioned a few times on HN. And https://render.com/ seems similar. For static websites those are free, I mean hard to compete with free. I'm sure they will add more features in the future, slowly becoming the next Heroku. We haven't done an official HN launch, but Render was built to solve the cost and inflexibility issues with Heroku. I'm the founder and always happy to answer questions. No-ops means someone else does the ops/system admin. Depending on requirements you may find a fully-automated or turnkey setup. I can deploy a WordPress site on Digital Ocean (and other platforms) with a few clicks. When something goes wrong you need ops support. Too many things can go wrong, I think impossible to anticipate all of them. The less familiar you (or whomever does system admin for you) is with the setup and environment the longer it will take to find and fix problems. The more complex the infrastructure (and serverless/microservices gets complex fast) the more expertise you need to deal with the inevitable problems. I don't think it's ever possible. The closest you can get is to pay someone else to do your ops, but then that's not really what you meant. The serverless framework gets you pretty dang close to no-ops though. You write your config, you write your code, you deploy and it just sort of works. Until you hit massive scale, and then you start needing to "operate" your serverless setup. Of course, serverless (the company) provides a commercial product to help you with that.