Analytical data at scale.
WHAT IT IS
A service for querying very large datasets using standard query language.
WHAT MAKES IT DIFFERENT
No infrastructure to provision, and querying happens across enormous data quickly.
HOW IT IS BILLED
By data processed by each query, or by reserved capacity.
WHY THAT MATTERS ENORMOUSLY
A careless query scanning an entire large table is expensive, every time it runs.
WHAT TO DO ABOUT THAT
Select only the columns you need Partition tables by date Cluster by frequently filtered columns Filter on the partition column
WHY SELECTING COLUMNS MATTERS
It is columnar storage, so unselected columns are not read or billed.
WHAT NEVER TO DO
Select all columns from a large table habitually.
WHAT TO SET
Maximum bytes billed per query, as a safeguard Budgets and alerts
WHAT TO USE FOR REPEATED QUERIES
Materialised views, or scheduled queries writing to a smaller table.
WHAT TO CHECK BEFORE RUNNING ANYTHING LARGE
The estimated bytes processed, which the tooling shows.
WHAT TO MONITOR
Cost per query and per user.
WHAT TO RESTRICT
Who may run queries against large datasets.
