[PHOENIX-4313] Use the bytes scanned to drive optimizer decisions

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Assignee: Unassigned
• Reporter: James R. Taylor
• Created: 20/Oct/17 21:16
• Updated: 20/Oct/17 21:16

Description

There's some low hanging fruit to optimize queries based on bytes scanned (i.e. StatementPlan.getEstimatedBytesToScan).

Sub-Tasks

• Base hash versus sort merge join decision on cost: Resolved
• Fallback to many-to-many join if hash join fails due to insufficient memory: Open
• Indexes not used when ordering by primary key: Resolved
• Use bytes scanned to drive join order: Open
• Refine cost for limit: Open