[DRILL-3270] Modify HbaseGroupScan to allow multiple fragments per region

• Type: Improvement
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 1.0.0
• Fix Version/s: Future
• Component/s: Query Planning & Optimization, Storage - HBase
• Assignee: Unassigned
• Reporter: Andy Pernsteiner
• Created: 10/Jun/15 15:56
• Updated: 01/Oct/15 16:36

Description

When performing a full HBASE or MapR-DB table scan using drill, it is observed within the resulting query profile that only one minor fragment is assigned per region, regardless of the size of the region. In the case of extremely large regions, especially if there are regions of mismatching sizes, this can result in poor performance and a low degree of parallelism.

One possible option (mentioned by sphillips) is to lazily compute the splits by assuming that the keys within a given region were evenly distributed (not perfect, but better than nothing), and perhaps have a 'max-frags-per-region' setting.