Caching Overview - Cube

Caching

Cube provides a powerful caching layer through pre-aggregations — materialized rollup tables that dramatically accelerate query performance and reduce data warehouse costs.

How it works

<Steps> <Step title="Define pre-aggregations"> Specify rollup tables in your data model with the measures and dimensions to cache. </Step> <Step title="Cube builds the cache"> Cube queries your data warehouse and stores results in Cube Store, a purpose-built caching engine. </Step> <Step title="Queries are served from cache"> When a query matches a pre-aggregation, Cube serves it from the cache instead of hitting the warehouse. </Step> <Step title="Automatic refresh"> Pre-aggregations are refreshed on a configurable schedule to keep data fresh. </Step> </Steps>

Defining pre-aggregations

Add pre-aggregations to your cube definitions:

yaml

cubes:
  - name: orders
    # ... measures and dimensions ...

    pre_aggregations:
      - name: orders_by_day
        measures:
          - count
          - total_amount
        dimensions:
          - status
        time_dimension: created_at
        granularity: day

Refresh strategy

Control how often pre-aggregations are rebuilt:

yaml

pre_aggregations:
  - name: orders_by_day
    measures:
      - count
    time_dimension: created_at
    granularity: day
    refresh_key:
      every: "1 hour"

<Warning> Setting very frequent refresh intervals can increase your data warehouse costs. Balance freshness with cost. </Warning>

Benefits

10-100x faster queries compared to hitting the warehouse directly
Reduced warehouse costs by minimizing direct queries
Consistent performance regardless of data volume
Automatic query routing — no changes needed in your application code