Hey everyone! I've been looking into how Apple's new container runtime stacks up against Docker Desktop, so I decided to run some benchmarks. I tested various factors including CPU performance, memory usage, disk I/O, and startup times. Here are my findings in a comparative table:
|Category|Docker|Apple|Units|
|:-|:-|:-|:-|
|CPU 1 thread|10939.81|11080.05|events/s|
|CPU all threads|53881.70|55415.57|events/s|
|Memory|81634.45|108588.00|MiB/s|
|Startup time|0.21|0.92|seconds|
If anyone's interested in the full charts and results, you can check them out here: [Full Benchmark](https://www.repoflow.io/blog/benchmarking-apple-containers-vs-docker-desktop). Let me know if you want me to run additional tests!
5 Answers
I’m particularly interested in file system performance, especially for small file I/O. Things like cloning a large Git repo or building a Node project using a directory shared from the host tend to show the biggest slowdowns for me, rather than CPU load.
Just to clarify, are you testing the new beta VMM of Docker or the older version?
This is all based on Apple's container runtime, which is fully independent. It utilizes OCI images launched into distinct VMs that are integrated properly.
That’s awesome you got it working! I struggled with it about a month ago when the OS was still in beta. Sounds like it’s worth another shot now.
Nice work! I’d love to see how non-native vs native architecture images perform. It could really highlight some differences!
Great idea! I hadn’t considered that. I’ll definitely run those tests and add the results to the benchmarks.
I’m also interested in a comparison to OrbStack. I personally don’t expect it to be as efficient as the Apple runtime.
Sure, I’ll include that in my next round of benchmarks.
I actually included I/O in my full benchmarks! It just didn’t fit into the post table.