[Hive] Hive Architecture

Hive Architecture

Driver

• fechtes the required APIs needed for the query that are modeled on JDBC & ODBC interfaces
• convert the Hive query into MR program from the help for the Compiliier

Compiler

• convert the Hive query into MR programs
• does semantic analysis of the program
• generates a execution plan with the help of a metastore

Metastore

• database storing all the structured information of the tables, partitions, number of columns in the tables, column data types, serializers, deserializers, etc
• by default Hive uses built-in Derby SQL server as metastore, not used in production since it provides single process storage(can’t run two simultaneous instances of Hive CLI)

Execution Engine

• connected with Hadoop
• executes a plan created by the compiler
• interact with the resourcemanager & namenode to fetch desired output data from the HDFS