[DRILL-6071] Limit batch size for flatten operator

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.12.0
• Fix Version/s: 1.13.0
• Component/s: Execution - Flow
• Labels: ready-to-commit
• Assignee: Padma Penumarthy
• Reporter: Padma Penumarthy
• Reviewer: Paul Rogers
• Created: 04/Jan/18 22:23
• Updated: 14/Mar/18 04:53
• Resolved: 31/Jan/18 01:44

Description

flatten currently uses an adaptive algorithm to control the outgoing batch size.
While processing the input batch, it adjusts the number of records in outgoing batch based on memory usage so far. Once memory usage exceeds the configured limit for a batch, the algorithm becomes more proactive and adjusts the limit half way through and end of every batch. All this periodic checking of memory usage is unnecessary overhead and impacts performance. Also, we will know only after the fact.

Instead, figure out how many rows should be there in the outgoing batch from incoming batch.
The way to do that would be to figure out average row size of the outgoing batch and based on that figure out how many rows can be there for a given amount of memory. value vectors provide us the necessary information to be able to figure this out.

Row count in output batch should be decided based on memory (with min 1 and max 64k rows) and not hard coded (to 4K) in code. Memory for output batch should be configurable system option.

Issue Links

• relates to: DRILL-6238 Batch sizing for operators
• links to: GitHub Pull Request #1091