Redshift Cheat Sheet

Overview

• A managed data warehouse
• Based on ParAccel solution, in turn based on PostgreSQL

Node Types

• DC2 = dense compute, SSD-backed, good for compute and I/O intensive workloads
• DS2 = dense storage, HDD-backed, good for huge dataset
• RA3 = good general purpose, recommended to replace DC2 and DS2
• It is odd that AWS does not have a page dedicated to node types, some info is buried in pricing page

Resizing

• Elastic resizing, can resize cluster and change node types, resizing done in minutes
• Classic resizing, resizing done in hours and days, not recommended anymore

Distribution Styles

• How data is distributed in cluster
• Styles
  • AUTO = automatically select from below strategies
  • ALL = full copy of data in every node, ok for smaller, not-updated-often tables
    • No significant performance gains for smaller table
  • EVEN = distribute in round-robin fashion, good if no joins
  • KEY = distribute by selected column, good when joining by those columns

COPY Command

• Use COPY command whenever you can, much more efficient than individual INSERTs
• Use a single COPY command, with multiple files, to enable parallel COPYing
  • AVOID using multiple commands
  • AVOID using a single file, split the file as a multiple of your slices