Batch send small blocks in ExchangeSender
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 1k
- Forks
- 423
- Avg merge
- 1d 15h
- Merged PRs (30d)
- 24
Description
For now we implemented a shuffle hash join in TiFlash MPP. In shuffle join, table will be partitioned by ranges. So there will be many fragmented data chunks (i.e. block with only tens or hundreds of rows).
We should find a way to batch the fragmented blocks, so that we can maximize locality and reduce overhead.
A list of MPP productions to refer:
- Impala
- Presto
- HAWQ
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start at ExchangeSender and study how fragmented blocks are produced by the shuffle hash join. Review the listed Impala, Presto, and HAWQ implementations for batching approaches, then define completion as batching small blocks while improving locality and reducing overhead.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- data-engineering, distributed-systems
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100