Impala and HBase

Impala is optimized for reading data batches, while HBase is key-value storage aimed for data look-ups
CREATE TABLE is not supported in Impala, it should be run from Hive; it maps new table to existing one in HBase
HBase key must be string
avoid HBase tables full scans - use the key for efficient querying
typical use-case is joining big table in Parquet with the small dimension table in HBase which also has WHERE condition with HBase key - advantage is that you can simply update that HBase table
another use-case: rapidly changing data tables
data ingestion: INSERT … VALUES is efficient, INSERT … SELECT, update can be simulated with successive INSERTs with the same key

Odkaz na článek

Vytvořil 28. dubna 2016 v 16:40:03 mira. Záznam nebyl nikdy upravován.

Diskuze ke článku