2 ms·

I'm curious about the remaining 20s to start the code on the container. What's the bottleneck there? It seems like you'd need to identify the container, tell it

I'm curious about the remaining 20s to start the code on the container. What's the bottleneck there? It seems like you'd need to identify the container, tell it which S3 object has the new source.pex, the container would download it, and then when you run `PEX_PATH=deps.pex ./source.pex` you're up. All that feels like it should take less than 20s. If picking the container takes long, you could probably start that process as sources.pex is built.

shalabhc 4y ago
Getting the message to the right container is one bottleneck. Currently this is routed through a couple of hops and includes some polling. This could all be optimized (if we had a direct line to the container) but the same messaging model is used in other contexts and would need architectural changes. Another bottleneck is running `source.pex` itself takes a few seconds to start up because it analyzes user code (and in some cases may do expensive computation.) But you're right: if `source.pex` as a hello world program, just downloading and running it should be pretty fast - I'd expect around 1s.