Haven't tested this locally, but seems like it might be a win. It's not really clear how much bandwidth CodeBuild machines should have, and we don't today log how much bandwidth we actually end up using for the uploads.
I suspect we may want a Rust-based impl that can give us more insight into the upload metrics if we end up optimizing this further. S3 operations are roughly ~12/60 minutes for a build today, so not entirely trivial, but not very long either.
Haven't tested this locally, but seems like it might be a win. It's not really clear how much bandwidth CodeBuild machines should have, and we don't today log how much bandwidth we actually end up using for the uploads.
I suspect we may want a Rust-based impl that can give us more insight into the upload metrics if we end up optimizing this further. S3 operations are roughly ~12/60 minutes for a build today, so not entirely trivial, but not very long either.