Query-first data modeling
Design a table for 'latest orders by user'. Pick partition key and clustering columns, then explain how your design avoids cross-partition scans.
Data / Knowledge library
Apache Cassandra
a distributed wide-column nosql database
Why learn Apache Cassandra?
Cassandra solves always-on, high-write, globally distributed data storage where horizontal scale, fault tolerance, and tunable consistency matter more than complex joins or strict single-node ACID behavior.
Where it came from
Created at Facebook by Avinash Lakshman and Prashant Malik (2008). Open-sourced by Facebook in 2008. Entered the Apache Incubator in 2009. Became an Apache Top-Level Project in 2010. Influenced by Amazon Dynamo (distribution) and Google Bigtable (data model).
- Created at Facebook by Avinash Lakshman and Prashant Malik (2008).
- Open-sourced by Facebook in 2008.
- Entered the Apache Incubator in 2009.
- Became an Apache Top-Level Project in 2010.
- Influenced by Amazon Dynamo (distribution) and Google Bigtable (data model).
Build your understanding
Work through these exercises in order. Try the task before opening a reference.
- Design a table for 'latest orders by user'. Pick partition key and clustering columns, then explain how your design avoids cross-partition scans.
- Run the same read/write flow with ONE, QUORUM, and ALL. Record latency and behavior after stopping one node. Summarize trade-offs in 5 bullets.
- Model IoT sensor events with TTL and day buckets. Implement insert + query for last 24 hours and explain hot-partition risk mitigations.
- Simulate node outage, recover it, then run validation steps (nodetool status, repair plan, read checks). Document what remained available and why.
- Compare primary-key query, materialized view, and SAI for one product-search scenario. Choose one and justify with expected read/write load.
More ways to practice
Consistency tuning drill
Run the same read/write flow with ONE, QUORUM, and ALL. Record latency and behavior after stopping one node. Summarize trade-offs in 5 bullets.
Time-series retention pattern
Model IoT sensor events with TTL and day buckets. Implement insert + query for last 24 hours and explain hot-partition risk mitigations.
Failure simulation and repair basics
Simulate node outage, recover it, then run validation steps (nodetool status, repair plan, read checks). Document what remained available and why.
Indexing decision memo
Compare primary-key query, materialized view, and SAI for one product-search scenario. Choose one and justify with expected read/write load.
Documentation and sources
- Apache Cassandra Docs (latest) (cassandra.apache.org)
- Apache Cassandra CQL Reference (cassandra.apache.org)
- Apache Cassandra Project Site (cassandra.apache.org)
- Apache Cassandra GitHub (github.com)
Catalog research notes
- Cassandra 5.x adds native vector type support for similarity-search use cases.
- Cassandra 5.x advances Storage-Attached Indexing (SAI) for scalable secondary indexing.
- Cassandra 4.1 introduced guardrails and better operational safety defaults.
- Cassandra 4.0 focused on major stability, observability, and performance improvements.