graphql-engine/dc-agents/guides/OLAP.md

Hasura + SQL (OLAP)

What do people like doing with SQL OLAP DBs?

• Use SQL to extract business insights from their data
  • SQL syntax and power is very important
• Since data is not frequently written and comes from another source consistency is assumed once data is in the OLAP store
  • Physical data modeling, normalization is not important
  • Joins are not important, data is often denormalized

The Hasura + X story:

• Hasura makes OLAP stores high-concurrency and low-latency
  • Query batching
  • Caching
• Hasura brings end-user authz to aggregations API (on single tables)
• Hasura brings out the best of OLAP features using PMs

What Hasura needs to have to make Hasura + OLAP CDWs awesome:

1. All single model features
2. Aggregations roadmap:
   • Group By
   • CTEs, Pivot tables
3. Parameterized models
4. High-concurrency by batching queries per second
5. Low-latency by allowing pre-warm caches
6. Solutions guide to ETL data to a row-storage engine instead of columnar
7. Async queries for slow running queries

Clickhouse specifics over and above:

1. Support dictionaries instead of joins
2. Support PKs and skip indices

Snowflake specifics over and above:

1. Support time based snapshots in parameterized models