Apache Cassandra
- 2. Agenda Cassandra basics Consistency Data modeling Real-world use cases Demo
- 5. A distributed database for managing large amounts of structured data across many commodity servers, while providing highly available service and no single point of failure Used by Apple, Comcast, Instagram, Spotify, eBay, Rackspace, Netflix What is Cassandra
- 6. Features Masterless architecture Fault tolerant Replication across nodes or data centers Linear scalability Tunable consistency
- 7. Tunable Consistency Replication factor Total number of replicas across the cluster Consistency level Number of replicas to respond before returning to the client Can be different for read and write ONE, QUORUM (RF/2 + 1), ALL
- 8. Scenario 1 RF = 3 Read = ONE Write = ONE INSERT
- 9. Scenario 1 RF = 3 Read = ONE Write = ONE Read
- 10. Scenario 2 RF = 3 Read = QUORUM Write = QUORUM INSERT
- 11. Scenario 2 RF = 3 Read = QUORUM Write = QUORUM Read
- 12. Strong consistency or Eventual consistency? If consistency is desired (nodes_written + nodes_read) > replication_factor Examples RF=3, Read=QUORUM, Write=QUORUM RF=3, Read=1, Write=ALL -> Optimized for read RF=3, Read=ALL, Write=1 -> Optimized for write
- 13. Scenario 3 RF = 3 Read = QUORUM Write = QUORUM INSERT
- 14. Scenario 3 RF = 3 Read = QUORUM Write = QUORUM Read
- 15. Data modeling
- 16. Compare to RDBMS No JOINs -> Prefer denormalization Model your data around the queries Limited transaction support
- 17. Key concepts Keyspace Similar to a schema in RDBMS Table Primary key = Partition key + Clustering column Partition key Defines the node on which the data is stored Clustering column Defines the order of data stored in a row
- 18. Phonebook A person has a name and a phone number Look up by name Look up by phone number
- 19. Create table CREATE TABLE person ( person_id uuid, name text, phone text, PRIMARY KEY (person_id) );
- 20. Create table CREATE TABLE person_by_phone ( name text, phone text, PRIMARY KEY (phone, name) );
- 21. Create table CREATE TABLE person_by_name ( name text, phone text, PRIMARY KEY (name, phone) );
- 22. Insert data INSERT INTO person (person_id, name, phone) VALUES (uuid(), 'Alice', '1000'); INSERT INTO person_by_name (name, phone) VALUES ('Alice', '1000'); INSERT INTO person_by_phone (name, phone) VALUES ('Alice', '1000'); INSERT INTO person (person_id, name, phone) VALUES (uuid(), 'Alice', '2000'); INSERT INTO person_by_name (name, phone) VALUES ('Alice', '2000'); INSERT INTO person_by_phone (name, phone) VALUES ('Alice', '2000');
- 23. Query by name select * from person_by_name where name='Alice'; name | phone -------+------- Alice | 1000 Alice | 2000
- 24. Query by phone select * from person_by_phone where phone='1000'; phone | name -------+------- 1000 | Alice
- 25. Demo
- 26. Real-world use cases Netflix recommendations Real-time data pipeline with Spark + Cassandra + Kafka https://www.youtube.com/watch?v=SxU0CJJ2nVE http://www.slideshare.net/DataStax/netflix-recommendations-using-spark-cassandra Multiple datacenters deployment at Uber With Apache Mesos https://www.youtube.com/watch?v=4Ap-1VT2ChU http://www.slideshare.net/DataStax/cassandra-on-mesos-across-multiple-datacenters-at- uber-abhishek-verma-c-summit-2016
- 28. Questions?