S3 download is quite slow
Hey all, I just learnt that my workers spent 4s to download total ~5mb files from s3. Is that normal? Or is the best practice to include these files in the runpod payload? If so, is there any size limit on the runpod post request? Thanks!
6 Replies
@ssssteven I do not expect this to be normal.
There is a limitation on the request sizes (10mb), and you have to encode it through the request body, not in like some form or something.
I have another endpoint that I do GET request against to download mp3 that is longer than 10mb, so surprised u are getting 4s
https://docs.runpod.io/serverless/references/operations
Endpoint operations | RunPod Documentation
Comprehensive guide on interacting with models using RunPod's API Endpoints without managing the pods yourself.
interesitng. Let me measure it with more details
thank you
are you just using requests to download the file from s3?
Only
/run
is limited to 10MB, /runsync
is limited to 20MB.
https://docs.runpod.io/serverless/workers/handlers/overview#development-and-deploymentOverview | RunPod Documentation
The function responsible for processing requests.
My requests are against a file storage server on Firebase (or GCP essentially under the hood)
@ssssteven The only thing I dont like about run-sync is that it can time out / also if anything goes wrong during the connection I just lose the result, vs /run at least you got a 30 min cache, so I try to just keep it under 10mb, so I have the flexibility of both
/runsync
doesn't time out. Its synchronous within 90s, then you either poll or use a webhook to get the result.ah
Great to know! Thank you @ashleyk ive learnt something new