Abstraction Layer on K8s, Thoughts? • Hacker News Board

Post by rtilwaniog » Thu, Mar 23, 2023, 7:40 AM UTC

We are a team of passionate DevOps and full-stack engineers who faced multiple issues while working because of the way developer teams function.

We started speaking to other engineers and everyone could relate to the struggle. Thus, in 2021, Humalect (https://humalect.com) was born.

Problems we saw: - DevOps is iterative, takes months to set up - Everyone is building their internal platform solution in-house to solve the same problem - Months spent setting up and maintaining open-source software and versioning - People are locked into various DevOps solutions and fear heavy migration debt to adopt better solutions - Things like security and scalability are an afterthought - DevOps teams are always overworked due to lack of visibility across various dev teams. The empathy towards DevOps teams and collaboration among multiple engineering teams needs to increase where everyone has visibility into infrastruture and release process.

This leads to

- Lower engineering velocity and reduced productivity - Compromised customer experience - Security issues - downtimes and leaks - Blown-up cloud costs - Heavy maintenance and tech debt

The proposed solution – Humalect

Platform knits best possible solutions for the most common DevOps scenarios.

Starting from infrastructure to application gets deployed to monitoring it and letting the user know if something goes down, all of it is pre-configured to run in the user's cloud account.

Secure and scalable infrastructure and application deployments delivered in minutes.

All user needs to do is to bring their code repositories and their cloud account.

Everything stays in user's cloud; no-lock in

The deliverables

- K8s clusters and stable environments created in your cloud account via auto generated IaC templates using best cloud and security practices - CI/CD enabled deployments directly from your repositories in a click - Built-in observability to help you bebug easily and keep track of resource usage - Auto scalability of infrastructure and your application guaranteed from day 1 - in-built support for creation of terraform, Dockerfile and YAML templates. - inbuilt logging and metrics for quick debugging, - integrations with leading tools for deeper insights - Help in saving upto 57% on your AWS EC2 workloads via reserved instances

It can be beneficial if user can relate to the points below:

Small team (upto 100 engineers): - your dev teams are spending a lot of time managing cloud/infrastruture - you have delayed sprints due to infra/env instability - you have to go live by a deadline or you have customers waiting but you don't seem to have a go live plan - You need a saviour to take you out of your misery

Large team (>100 engineers) - You have a lot of tech debt/lock-in due to your past stack/infra/cloud choices and now you are looking to move to latest technologies available - You recognise that change is needed - You recognise that upskilling for the needed migration is difficult and time taking given current responsibilities of your team - It will take you months to build the solution in-house and the onus of maintenance relies on you post creation - You know the productivity can be better or there can be cost savings if you need to newer approaches but you don't seem to figure out an execution plan.

Benefits our users have seen

- Better visibility into infra and deployment lifecycles - faster release cycles, thanks to time saved on broken environments/infra management - better cost control due to auto scalability and better cloud cost control - better collaboration among teams as siloing goes away - better developer productivity - scalable infrastructure and full visibility via integrations

Are we thinking straight? Comments? feedback?

Post by igorzij » Thu, Mar 23, 2023, 5:14 PM UTC

Hey friends, Digger.dev founder here. I BEG YOU - let this dream go if you want your co to still be still around in a few years. We've spent 2 years chasing it. So did 20+ other passionate teams. Look at this list: https://diggerdev.notion.site/DX-platforms-for-your-cloud-ac...

It took us 2 years to realise one simple thing: it's not wrong, it's just not the right starting point.

Indeed this kind of platform that'd make cloud ops much easier, decouple dev from ops, self-service and everything, on top of Terraform, is inevitable. It makes perfect sense!

But people don't need platforms. People need great solutions to at-hand problems. And for such a platform to take off, it needs to be best in class at _every single thing_ it does - everything from infrastructure provisioning to ci / cd to observability. No team in the world has capacity to execute at such scale from scratch, not even Microsoft (owning both Github and Azure).

Worse still, all these problems occur at different times in customer's company lifecycle, and often to different people in their org structure. This makes it next to impossible to narrow down to a use case or specific customer persona. This feels like a force that constantly tries to tear your startup apart. I'm sure you know what I'm talking about.

I think that the only way to make a horizontal platform is incremental expansion of independently successful vertical parts. Just having "integrations" with various third parties is not enough - for every problem your platform solving itself, there's likely a specialised tool that does it better. So there's got to be _one thing_ that you do better than everyone else. Just one. Unless you have it, people won't want your platform. And when you have it, you might just as well not have anything else because none of the other stuff would matter.

Find the one thing that people struggle with, build a simple solution, and make sure you are the best in the world in this one little thing. That's not hard to do, but extremely hard to truly believe in deep down. But that's the only way.

We have recently found our _one thing_: CI for Terraform. Digger makes it easy to run terraform in Github Actions. That's it, that's all we do.

Good news: this does not invalidate the big platform vision in the future - you can still build up towards it incrementally from your small thing! If anything having tried that is a competitive advantage against clueless incumbents.

I also may be wrong. Prior to Digger I was working on Palantir Foundry which faced a similar "backwards GTM" challenge - so I have 2 biased data points. But the folks at Qovery seem to be doubling down on the platform route. I'm rooting for them as well as for everyone else on the list. With enough perseverance and capital it might well be possible to pull it off, especially via top-down sales-led enterprise route. Someone will eventually make such a platform because it's inevitable.