Skip to content

Conversation

@andygrove
Copy link
Member

@andygrove andygrove commented Nov 17, 2024

Follows on from #47

Closes #46

Description

  • Shuffle files are still written to local disk in ShuffleWriterExec, but they are then uploaded to object storage
  • ShuffleReaderExec then downloads the shuffle files from object storage and deletes them

final_stage_id = graph.get_final_query_stage().id()
partitions = schedule_execution(graph, final_stage_id, True)
# serialize the query stages and store in Ray object store
query_stages = [
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am a little puzzled, are these the query_stages or the serialized execution plans?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

these are the serialized execution plans for each query stage:

graph.get_query_stage(i).get_execution_plan_bytes()

@andygrove andygrove changed the title feat: Add support for object storage based shuffle feat: Add support for cloud object storage (e.g. s3) based shuffle Nov 18, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Fix scalability limitations of current implementation

2 participants