Tanuj Khurana created PHOENIX-7319:
--------------------------------------
Summary: Leverage Bloom Filters to improve performance on write
path
Key: PHOENIX-7319
URL: https://issues.apache.org/jira/browse/PHOENIX-7319
Project: Phoenix
Issue Type: Improvement
Affects Versions: 5.2.0
Reporter: Tanuj Khurana
Assignee: Tanuj Khurana
Fix For: 5.2.1, 5.3.0
On the write path if the write is an atomic upsert or if the table has one or
more indexes Phoenix first does a read. All these reads on the data table are
point lookups. Bloom Filters can help optimize the performance of these
lookups.
* For new rows (inserts), the point lookup will not return any result. This
negative lookup is ideal for bloom filters as our read can return by just
checking the bloom filter block.
* For updates, since new updates get accumulated into memstore and then
flushed into new store files. A region can have multiple store files and when
doing a read we have to read multiple store files. Bloom filter can help
eliminate which store files should be read.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
