Stores data on a single node

Avoids failures

Uses replication only

Uses lineage to recompute lost data

Relational Database Data

Reduced Data Domain

Random Data Division

Resilient Distributed Dataset

collect()

flatMap()

reduceByKey()

map()

flatMap()

map()

filter()

reduce()

RDD

Driver

Executor

Cluster Manager

Uses disk storage only

In-memory processing for faster computation

Only supports SQL

No need for cluster

Only numeric data

Strict relational only

Nested and semi-structured data

Flat file only

Faster hardware usage

Automatic query optimization

Uses SQL

No need for Hadoop

Groups data

Transforms and projects data

Filters rows

Joins datasets

JOIN

FOREACH

GROUP

FILTER