MongoDB and Drupal - Drupal in LA Feb 2012

MongoDB and Drupal presented to the Los Angeles High Performance Meetup group in February 2012

Transcript

1. Drupal and MongoDB High Performance Meetup - LA February 2012 BTMash
2. Agenda • RDBMS: Issues • What is MongoDB? ◦ About NoSQL / Document storage ◦ Queries • Why MongoDB? ◦ Pros ◦ Cons • What about Drupal? ◦ MongoDB module ◦ Field Queries • Why Drupal + Mongo? ◦ Pros • Setup + Demo • Q&A
3. Relational DB Issues Not all that much. HOWEVER: • Once you have related data, you need to join them. • Sometimes, indexing doesn't work very well. ◦ Introduce denormalization. ◦ Build extra tables. • Changes can lock data. ◦ Downtime ◦ Load on system • Code changes are not automatically coupled with schema changes. ◦ Coordination can be difficult.
4. What is MongoDB? About MongoDB • Document oriented NoSQL Database System ◦ Written in C++ ◦ Released in 2009 Query Results looks like JSON { "v" : 1, "key" : { "_id" : 1 }, "ns" : "db.fields_current.search_api_index", "name" : "_id_" } ...
5. What is MongoDB (cont'd) • Databases consist of collections ◦ Collections contains objects. • CLI Queries are somewhat similar to sql (except not): db.fields_current.node.find({'uid': 0, 'field_field_name.value': 1}); db_fields_current.node.find({'_id': 100}); db.fields_current.node.find({'title': /about/i}, {title: true}).limit(4); db_fields_current.node.save(JSONified Node Object); • Will update if the object id already exists. db_fields_current.node.remove({my_json_parameters});
6. Why MongoDB? Pros • Highly Available • Easily Scalable • Partition Tolerant • Tableless / Schemaless • Very Fast Example Whitehouse.gov direct engagement • 15k/day requests • 2M records in db • Replication issues since db was nearly 4gb. • Switch to MongoDB ◦ Relatively easy replication / sharding • Now over 180M collections in collection.
7. Why MongoDB? Cons • Not exactly query-less. • Still have to format your query to it correctly. • Range query / sort must be based on last column in index. • Have a limit of 64 indexes. • Cannot join across entities (and thus filter on data from multiple entities). • No DBTNG Implementation.
8. MongoDB Module http://drupal.org/project/mongodb • mongodb: core support library for other modules • mongodb_block: Store blocks in mongo • mongodb_cache: Cache using mongo • mongodb_session: store sessions in mongo • mongodb_watchdog: Store watchdog in mongo • mongodb_queue: Store queues in mongodb • mongodb_field_storage: Store fields in mongodb
9. MongoDB core $collection = mongodb_collection($collection_name) $collection->findOne(array('_id' => (string)$nid)); $collection->find($conditions); $collection->remove(array('_type' => 'blah')); $collection->save(array($sanitized_node)); That's really it ^_^
10. MongoDB Watchdog • Lower strain on db server. • Useful when there might be a lot of activity on site (be it good or bad). • Can cap collection size. • Quick purge. • Enable it and you're ready to go (disable dblog - you don't need it).
11. MongoDB Cache / Sessions $conf['cache_backends'][] = 'sites/all/modules/mongodb/mongodb_cache/mongodb_cache.inc'; $conf['cache_default_class'] = 'DrupalMongoDBCache'; $conf['session_inc'] = 'sites/all/modules/mongodb/mongodb_session/mongodb_session.inc'; • Sessions will likely be helpful but cache...not sure what kind of difference is to be seen there. Better off placing efforts in using apc / memcache / something else.
12. MongoDB Field Storage • Enable MongoDB Field Storage module. • $conf['field_storage_default'] = 'mongodb_field_storage'; • Biggest gains to site are from this. PROS • Faster saving, deletion, node retrieval. • Fully compatible with entityfieldquery. • No real limits on field lengths (*no image alt length issues) • Partial Views Integration through EntityFieldQueryViews (http://drupal.org/project/efq_views) - could use some help
13. Demo
14. Questions? Thank you :)