VEED’s engineering team was dealing with longer and longer CI times on GitHub Actions, resulting in decreased productivity for the whole team.
Now, VEED’s engineers have cut their CI times in half and their costs by 70%, which has been an insane productivity boost for the whole team.
VEED is an AI-powered online video creation platform that makes video creation easy and accessible for everyone. And by everyone, I really do mean everyone. It's no surprise they’re trusted by 76% of Fortune 500 companies.
With that much of the Fortune 500 at the tail end of their CI/CD pipeline, every pull request and production deployment undergoes a huge set of validations that give the VEED team the confidence needed to merge and release new features and fixes, fast.
The Problem
When your local machine is faster than Github.
Before switching to Blacksmith, VEED relied on GitHub-hosted runners for their CI. Back then, they were facing painfully long CI times, thanks to the super low clock speed of GitHub’s CPUs. Máté Nagy, Software Engineer at VEED, was having an especially rough time when what took him 10 minutes to run locally ended up taking a whopping 22 minutes on GItHub-hosted runners.
This was not going to cut it. VEED needed speed—especially for their End-to-End (E2E) tests, which were running Playwright and consuming a lot of compute resources. And to make matters worse, their tests were interacting with a server in the EU, but GitHub-hosted runners are all located in the US, which seriously increases latency for the team’s tests.
On top of everything, and here comes the broken record again, VEED was also hit by frequent GitHub Action outages, causing jobs to not start at all. It was a perfect storm!
GitHub was really getting out of hand toward the last year with the outages.
— Máté Nagy, Staff Software Engineer, VEED.
At that point, enough was enough. Given the significant hit to productivity this was causing, the team decided it was high time to find the fastest, most powerful CI they could get their hands on.
The Solution
Selecting the fastest option.
To begin with, VEED wanted a drop-in replacement for GitHub-hosted runners to make the migration as painless as possible. They’ve had enough headaches with CI, after all.
The moment Máté discovered us, he was sold—especially because we offer the most powerful CPUs commercially available. Seriously, at the time, nothing else came close. In fact, the CPU we offered was the same one he had in his own local machine!
Given Máté’s focus on performance, he was motivated in making sure VEED’s CI would always be on the most powerful CPUs available. Now, if we were on AWS, we’d be stuck waiting for AWS to add new, powerful CPUs to their instance offerings whenever a fresh chip hit the market. But with our bare-metal fleet, we’re always able to grab the latest and greatest in CPU performance as soon as it’s available. In fact, we’re already in the process of procuring Intel’s Arrow Lake chips, which will boost our customers’ CI performance by another 20%! Our commitment to being the fastest way to run your GitHub Actions in the long run was exactly what Máté was looking for.
Moreover, the fact that we had a fleet based in Europe meant that network latency would drop significantly for their tests, especially when communicating with their EU servers. Máté was also happy to see our drop-in replacements for GitHub’s caching actions because, well, let’s just say we’ve patched some quirky GitHub behaviors.
All of the fun performance bits, plus the promise of cost reductions for his manager, was enough to convince Máté and the team to bet on Blacksmith!
Running our CI on the most powerful CPUs available commercially, the same ones I use locally, is awesome.
— Máté Nagy, Staff Software Engineer, VEED.
The Results
From 28 to 14 minutes waiting time on a PR.
Today, VEED runs their CI faster than ever before. More than 2x faster. A PR that used to take a sluggish 28 minutes to pass all checks now wraps up in just 14-18 minutes. And the cherry on top? Switching to Blacksmith slashed their CI costs by a whopping 70%.
This newfound speed and cost efficiency even allowed them to justify paying for a bigger runner in certain use cases—something that wouldn’t have made financial sense on GitHub-hosted runners. This little upgrade gave some of their test suites an additional 15-20% performance boost. Talk about a win-win!
In tech there's almost always a tradeoff. When improving one metric, another one gets worse. Moving to Blacksmith was one of those rare success moments, where I saw no tradeoffs, only wins.
— Máté Nagy, Staff Software Engineer, VEED.
If you’re currently looking for the fastest and most reliable way to run your GitHub Actions, look no further. We got you covered.
