Show HN: DeltaCI: Bare Metal CI/CD with much faster builds
In this technical retrospective (2021), @thinkafterbef shares practical architectural lessons learned from running, scaling, or migrating production software under lean constraints.
Core Principles for Independent Builders:
Defaulting to simpler architectures (such as monolithic frameworks, SQLite, or managed VPS) prevents early engineering bankruptcy.
Evaluating true compute and egress costs reveals that high cloud bills can often be slashed by 80%+ through judicious re-platforming.
Code you write at 2 AM as a solo engineer must be easy to debug six months later without complex microservice dependencies.
Community Debate Themes:
HN engineers debated the tradeoffs between cloud convenience and raw server economics, database indexing strategies, and resisting premature architectural complexity.
CATEGORY Frugal Engineering →
HISTORICAL ERA Pandemic Bubble (2020–2022)
COMMUNITY SCORE ★ 17 points
TOTAL DISCUSSIONS 5 comments
Hello HN, we’re launching DeltaCI, a cloud-hosted CI built with self-hosted CI performance.
In our previous day jobs, we sometimes see ourselves keep hitting refresh on GitHub for 10 minutes just to wait for the test to finish then we can merge a PR.
In every build run, docker build image has to be pulled, `npm ci` has to run although we haven't touched package.json. Even with an optimized caching setup, it's still painfully slow to pull a ton of node_modules from cache.
Existing CIs also run their jobs on a Cloud VM which hardly go beyond 3 GHz, but our dev laptop can easily boost and 5GHz, then finish the same build in half of the time.
We always migrated to a self hosted CI or build agent once the problem went beyond our tolerance. However it was always extra effort to set it up and maintain it.
Since all existing solutions charge you on a per-minute basis, they literally have no incentive to make it faster for us. In the end, less build minutes = less revenue for them.
So we came up with the idea of inventing a new cloud hosted CI to solve all these.
Instead of paying AWS for the over priced VMs, we run your build jobs on latest Zen2, Zen3, 10th gen Intel desktop CPUs, as well as EPYC server cpus with upto 64 vCPU and 128G of RAM if you have builds which can make use of that.
We developed our own VM hypervisor management solution based on libvirt. It will run your builds on the same disk image which are kept persistent between job runs, just like your local dev machine. We also made it easy for you to avoid repeating steps like installing dependency. For example, you can easily do this:
steps:
``` - run: sudo apt install rsync when: first_run - run: npm ci when.changed: "\*/package\*.json"
``` Our system keeps your build environment persistent, and it also takes care of that rsync will be installed the first time you run the job, node_modules will only need to be touched when you actually have a dependency change. If you are building Docker images in your CI, then you also get Docker Layer Caching for free, since the building environment is persistent.
We dog-fooded this setup for half a year, and it cut off our web applications build time of testing/e2e testing/releasing from 12 minutes to just under 4 minutes. We also benchmarked several open source projects, and also saw similar speedup. You find them at https://deltaci.com/ and check out the actual job runs.
It's the first time we ever show this early stage product to the public, and it still has rough edges to be smoothed out. So we prepared a coupon code HACKERNEWS that’ll give you 99% off for 3 months.
We’d love to hear your thoughts.