drupalcon denver 2012 - steve francia mongodb

2012dcdenver-stevefrancia-mongodb.txt

mongodb  - steve francia from 10gen (new york co)

uses native storage methods
useful for large storage 

example blog 
- mongo db using php
- mongodb doesn't need to convert to array
- just pass to connection object with array to collection object
- memcache -> query one dimension
- multi dimension and multivalue

Problems with current databases
- converting database from arrays 
- no need to convert
- learning curve is issue
- no need for joins

Nice features
- support ad hoc query - no need to require indexes
- one document
- query language is intuitive
- no need for external connectors

pain points
- rarely need to use joins
- no foreign key integrity
- doesn't support 
- sql is great for aggregation - sql has methods add etc

pipeline
- bunch of simple things
- notion of using command line
- e.g. pipeline method -> grep | etc
- used for aggregation methods
- don't use group
- aggregation framework runs faster than map produce

Cassandra vs mongodb
- useful for loggin - bread and butter
- cassandra optimized for insert
- cassandra used in parralel with other technologies
- several use cases

couchdb vs mongodb
- doesn't have sharding
- ubuntu -> blog post about not using couch
- not used for scaling
- general purpose database

indexes
- memcache wouldn't give performance boost on query
- mongodb uses indexes as well; same benefits
- don't query from index - takes longer
- with mdb you can use disk
- mdb allows you to index arrays of values

mongodb and scalability
- doubleclick displays lots of ads and record all images and clicks
- sharding natively - replication provides high availability
- possible to go single node to sharded without changing application

100% no room to allow it. must have memory to be able to do this
- 80% peak good part to provide read 
- determine shard key