Performance experiement : Hive vs Redshift

Performance experiement : Hive vs Redshift

This experiment was done by Airbnb,


Experiment Setup:

Redshift — 16 node cluster ($13.60 per hour — $0.85 per hour per node) Node Type: dw.hs1.xlarge CPU: 4.4 EC2 Compute Units (2 virtual cores) per node Memory: 15 GiB per node I/O Performance: Moderate Platform: 64-bit Storage: 3 HDD with 2 TB of storage per node

Hive/EMR — 44 node cluster ($57 per hour) Node Type: Cluster Compute Quadruple Extra Large Memory: 23.00 GB CPU: 33.5 (2xIntel Xeon X5570) Storage: 1690 GB (2×840 GB) Platform: 64-bit I/O Performance: Very High

Test 1: