$3000 Data Warehouse — Redshift vs. Postgres
Whether it’s 30-day retention or unique sessions, many analytics queries rely on being able to count the distinct number of elements in a set very fast. On average, Redshift was 200x faster than RDS Postgres for these queries.
..[1] Postgres stores data by row. This means you have to read the whole table to sum the price column.
Redshift stores its data organized by column. This allows the database to compress records because they’re all the same data type. Once they’re compressed, there’s less data to read off disk and store in RAM
.. [2] Postgres does not use multiple cores for a single query. While this allows more queries to run in parallel, no single query can use all of the machine’s resources.
.. [3] A Redshift cluster can achieve higher much IOPS. Each node reads from a different disk and the IOPS sums across the cluster. Our benchmark cluster achieved over 50K IOPS.