[ASTERIXDB-1639] Need a spatial test case with a high point dup factor

• Type: Test
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: *DB - AsterixDB, HYR - Hyracks, STO - Storage
• Labels: soon
• Assignee: Taewoo Kim
• Reporter: Michael J. Carey
• Created: 10/Sep/16 02:39
• Updated: 10/Sep/16 18:45

Description

We need a few LSM R-tree test cases where we have many leaves worth of data (which could be achieved by making an artificially small NC config?) that have the same key - to make sure that we can handle that case properly. (I'm wondering after talking with Wail if that's the root of his problems a few weeks ago - he had a high duplicate rate.) E.g., we should trying to insert a ton of data all at one of the same 2-3 unique spatial points. It would be good for there to be enough data that multi-level Hilbert sorting is required as well. This is likely to be a time-consuming test so it should be in our period (not per-checkin) tests. We should actually do this extreme-dup-case test for all index types, but R trees are suspected of maybe doing this wrong. Who would be best to write/run this test w/o much effort?