Script mode

MapReduce mode

Standalone mode

Local mode

Real-time processing

Writing complex MapReduce programs using scripting

Managing HDFS storage

Creating databases

Requires too much memory

It lacks SQL support

High latency due to batch processing

Cannot process large data

ORDER BY

GROUP BY

WHERE

HAVING

Eliminates need for indexing

Reduces storage size

Improves query performance by scanning less data

Increases redundancy

Temporary table

Managed table

External table

View

Executed as Python code

Direct execution on RAM

Converted into MapReduce/Spark jobs

Sent to HBase

Managing HDFS blocks

Querying large datasets using SQL-like language

Storing data in memory

Real-time data processing

No support for queries

Distributed storage without relational structure

Lack of indexing

High memory usage

FETCH

SELECT

READ

GET