Conversation
|
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
c84c8bc to
13b0aa9
Compare
13b0aa9 to
441623c
Compare
6224574 to
1af51b7
Compare
📊 Benchmark ComparisonCross-matrix comparison of workflow performance across frameworks and backends. workflow with no steps
workflow with 1 step
workflow with 10 sequential steps
workflow with 10 parallel steps
Summary: Fastest Framework by Backend
Summary: Fastest Backend by Framework
Column Definitions
Backends:
|
1af51b7 to
a9858e9
Compare
VaguelySerious
left a comment
There was a problem hiding this comment.
LGTM. The only thing I dislike is having a CI job that runs 15 minute (darn postgres). I wonder if there's something trivial we can do to speed that up to >=5 minutes? Just so it doesn't pollute CI too much. But feel free to merge either way and we'll speed up later.
It'd also be really cool if the bench workflows/steps could console log their ID + timestamps, so that it's trivial to check the CI logs to see where/how the runs/steps are slow. Should be a tiny change, can also do separately

Add some simple vitest based performance benchmarking for local world, postgres world, and vercel world, along with a github action to track it
Every PR will now run benchmarks across a few frameworks and all worlds. Once complete, a summary is posted with the results: https://github.com/vercel/workflow/actions/runs/19782846073/attempts/1#summary-56685707991