From Mac-Only to Linux Powerhouse: Journey to #1 CI Performance¶
Cirrus Runners has long been known for its blazing-fast macOS runners (thanks to Apple Silicon), but over time we’ve quietly built an equally powerful solution for Linux workloads. While our service initially launched with a focus on mobile CI (macOS and iOS builds), today a large chunk of our customers use Cirrus Runners exclusively for Linux builds. In other words, we’re not just a Mac runner provider – we’re a Linux CI powerhouse as well.
This post outlines the path we took to get there and how our vertically integrated approach (we control hardware, networking, and virtualization) lets us squeeze out every bit of performance.
From Tart to Vetu: Bridging Apple VMs to Linux Servers¶
Our journey began with Tart, our source-available virtualization tool for Apple Silicon. Initially, Tart was built to spin up macOS VMs on M1 Macs for CI/CD, delivering near-native speeds using Apple’s Virtualization.Framework. But demand quickly expanded—Tart users wanted Linux guests, and Cirrus Runners customers running iOS builds started spreading the word to their Android counterparts, who were also hungry for faster, more reliable infrastructure.
We answered by adding Linux guest support to Tart and, to maximize flexibility, enabling x86_64 VM image support. This meant we could build and run Linux VMs on both Apple Silicon and Intel-based Apple hardware. By mid-2023, you could create a Linux VM—ARM or x86—on a Mac and drop it straight into automation. This cross-architecture capability was a turning point: it let us reuse Apple-built machine images and run them on commodity Linux servers.
That’s where Vetu comes in. Inspired by Tart, Vetu is a runtime designed to run Tart-built Linux VMs on Linux hosts, whether they’re x86 or Arm. A VM created on an Apple Silicon or Intel Mac can be deployed on a high-performance Linux server with zero rebuilds. Under the hood, we chose Cloud Hypervisor over Firecracker to unlock advanced features like GPU passthrough. Beyond features, Cloud Hypervisor’s modern‑minimal design exposes richer CPU capabilities (e.g., AVX‑512) and enables NUMA/affinity tuning and vCPU hot‑plug—delivering better single‑thread performance than Firecracker’s intentionally narrow CPUID and fixed‑vCPU model.
With Tart handling VM creation and Vetu handling cross-platform execution, we built a seamless pipeline for delivering high-performance Linux CI environments anywhere.
Vertical Integration = Unmatched Performance¶
Fast-forward to 2025: we recently upgraded all our infrastructure for running Linux VMs, fully harnessing this Tart+Vetu stack on our own hardware. The results have been outstanding – Cirrus Runners is now the #1 performance-rated provider for GitHub Actions runners on x64 machines, according to an independent benchmark concluded by RunsOn. This means faster build and test times for your workflows.
Our vertical integration plays a huge role here: because we control the stack end-to-end – from selecting high-frequency CPUs (e.g. AMD Ryzen 7950X3D on our hosts) to networking, storage, and our custom virtualization layer – we can optimize everything for speed and reliability. According to our internal comparisons, Cirrus Runners delivers 2–3× better performance than standard GitHub-hosted runners, and at a fraction of the cost.
Equally important is consistency – by owning the hardware and not relying on multi-tenant cloud VMs, we avoid the “noisy neighbor” problem. Each runner VM is backed by dedicated resources on our servers, giving predictable performance run after run. And because we manage our own networking gear in the data center, the latency between runners and artifact storage or caches is low. In short, vertical integration lets us push the envelope on raw speed while also ensuring stability.
Rock-Solid Reliability and Customer Trust¶
Performance isn’t our only strength – we know enterprise customers also demand security, compliance, and uptime. Cirrus Runners is SOC 2 Type II compliant, reflecting our commitment to strong security practices and operational maturity. We’ve earned the trust of many large companies (including ones we can’t publicly name) by meeting their rigorous compliance requirements. And for the record, our uptime has been exemplary. We’ve built redundancy and monitoring at every layer of our service. In fact, Zendesk Engineering team noted they experienced “no downtime in the past year” of using Cirrus Runners.
Cirrus Runners provide a remarkable speed increase with their M4 Pro runners, significantly reducing build times. They offer a cost-effective solution that enhances productivity while minimizing expenses. In addition to these benefits, we experienced no downtime throughout the past year, making the service very reliable. Highly recommended!
(“M4 Pro” refers to one of our high-performance Mac runner types – even though this quote is about macOS builds, the same reliability and speed apply to our Linux offerings!)
Many organizations choose Cirrus Runners not just for speed, but for peace of mind and cost control. Our pricing is a simple fixed monthly rate with unlimited minutes per concurrent runner – meaning you can run as many CI jobs as you need without worrying about runaway costs. There’s no metered billing by the minute or egress data transferred; every runner is a predictable expense. This model, combined with zero infrastructure maintenance on the customer’s end, is incredibly appealing to DevOps teams who might otherwise spend time babysitting their own build servers or juggling cloud credits.
Case in point: the Bitcoin open-source project. Bitcoin Core had been maintaining its own CI hardware for their Linux builds – an expensive and effort-intensive endeavor for such a widely watched project. Just this week, they switched to Cirrus Runners’ managed Linux infrastructure for their GitHub Actions CI. The move gives their developers faster turnaround on tests and frees the team from worrying about server upkeep. The Bitcoin maintainers were particularly drawn by our flat pricing for unlimited usage and the elimination of on-prem maintenance, which lets them focus on development instead of DevOps overhead. And they did a thorough research:
Conclusion: Linux Builds at Full Throttle¶
Cirrus Runners may have started in the Apple world, but today it’s a cross-platform CI engine that can turbocharge your Linux workflows just as well as your Mac jobs. By innovating on virtualization (Tart and Vetu), embracing modern hypervisors, and tightly integrating every layer of our stack, we’ve created a Linux CI platform that’s both exceptionally fast and simple to use.
DevOps engineers looking for a GitHub Actions alternative will find that with Cirrus Runners they get the best of all worlds: top-tier performance, predictable pricing, enterprise-grade security, and a fully managed experience. We handle the heavy lifting – from hardware to hypervisor – so you can just specify your workflow and let it run at full throttle.
If you’ve thought of Cirrus Runners as “the Mac CI service,” it’s time to take another look. Whether you need to speed up Android builds, Docker image pipelines, kernel compilations, or even GPU-accelerated ML tests, our Linux runners are ready to go. We’re excited about the road ahead and we invite you to give our Linux infrastructure a try. Your builds (and your finance team) will thank you! 🚀