NoSQL
5 likes1,133 views
The document discusses various types of NoSQL non-relational data stores, highlighting key-value stores like Redis, document stores such as MongoDB and CouchDB, and extensible record stores like Bigtable and Cassandra. It emphasizes features like scalability, eventual consistency, and the differences in architecture among these systems. The author concludes by sharing their practical experience with NoSQL databases and their choices based on specific application needs.
NoSQL
- 1. NoSQL
- 2. Non-relational data stores Eric Evans, Rackspace, early 2009
- 3. Not only SQL No to SQL
- 5. Giving up ACID Eventual consistency Schemaless Scale horizontally
- 6. ACID Atomicity Consistency Isolation Durability
- 7. BASE Basically Available Soft state Eventually consistent
- 8. 1. Key Value stores
- 9. Redis C in memory journal append-only file list/set/sorted set boolean operation
- 10. Erlang similar to Redis in memory EU sponsored Scalaris
- 11. Mixi (Facebook japonais) Tokyo Tyrant (client) Tokyo Cabinet (server) several server mecanism asynchronous replication multi-master, master/slave map/reduce in Lua API compatibility w Memcached in memory or disk Tokyo Tyrant
- 12. Java LinkedIn in memory pluggable storage Project Voldemort
- 13. Erlang REST JSON in memory or disk Riak
- 15. Amazon Web Services proprietary pay as you go SimpleDB
- 16. Erlang Apache map/reduce in Javascript REST JSON CouchDB
- 17. C++ 10gen query & map/reduce in Javascript socket BSON (Binary JSON) automatic sharding GridFS master/slave asynchronous replication MongoDB
- 18. 3. Extensible record stores
- 19. ” Bigtable: A Distributed Storage System for Structured Data”
- 20. Java Apache BigTable w Hadoop (DFS) HBase
- 21. C++ Zvents, Baidu very similar to Hbase needs DFS (like Hadoop) HQL HyperTable
- 22. Java Facebook then Apache very similar to Hbase marriage of Dynamo and BigTable supercolumn, column tunable concistency decentralized fault tolerant (data center) elasticity Cassandra
- 23. For who ?
- 24. ” If you're asking me, should I use a NoSQL store ? Then the answer is no.”
- 25. !CouchDB MongoDB Redis (?) Our choice @novelys
- 26. pretty straightforward pretty close from our previous DB drivers developped in parallel several ODM in Ruby schema-less ! easy querying no fulltext search (well not really) rethinking how we store data kind of builtin denormalization Experience w Mongo http://www.rocketblogging.com