Redis in Practice

PHP-IST Istanbul, 2014

Transcript

1. Redis In Practice PHP-IST 2014, Osman Ungur
2. What is Redis • In-memory persistent key-value store • Hold all the data in memory • Write that out to disk for persistence • Much more than a simple key-value store
3. What is Redis • Created by Salvatore Sanfilippo • Written in C, March 2009 • BSD licensed (free and open) • Sponsored by Pivotal, since May 2013
4. Notable features • Transactions • Atomic operations • Expiry • Single threaded, no locking necessary
5. Notable features • Lua scripting • Publish / Subscribe support • Event driven • Amazing performance!
6. Data structures • Specific data structures for specific problems • Exposes five different data structures • Only one of which is a typical key-value structure • Other types is lists, sets and hashes
7. Keys to understanding Redis • Understanding these five data structures • How they work • What methods they expose • What you can model with them
8. Databases • Contains a set of data • Databases are simply identified by a number • Group all of an application’s data together • Keep it separate from another application’s
9. Keys and Values • Keys are how you identify pieces of data • Example: the_answer_to_life, it's binary safe • Values represent the actual data associated with the key • They can be anything, strings, integers, serialized objects, binary data
10. Commands COMMAND parameters ! SET the_answer_to_life 42 OK GET the_answer_to_life 42
11. Querying ! • Keys are everything and values are nothing • Redis doesn’t allow you to query an object’s values • We can’t find "the users which is living in Istanbul" • This is why values can be anything
12. Memory and Persistence • Redis keeps all your data in memory • RAM is still the most expensive part of server hardware • Snapshots the database to disk based on how many keys have changed • Alternatively or in addition to snapshotting, it can run in append mode
13. What isn't Redis • RDBMS (mySQL, postgreSQL) • Column (Cassandra, HBase) • Document (MongoDB, Couchbase) • Graph (OrientDB, Neo4J)
14. Installation apt-get install redis-server yum install redis brew install redis Or you can compile it from source
15. Using with PHP • Predis, mature and supported (predis/predis) • SncRedisBundle, bundle for Symfony2 (snc/redis- bundle) • phpredis, written in C as a PHP module • Packages for Laravel, Nette and Doctrine Redis Cache
16. Strings • Most basic kind of Redis value • Binary safe, can store any data • Maximum 512 Megabytes length • Useful for output caches, counters
17. Strings • Atomic increments and decrements with float support • Bitwise operations • String manipulation, Expiration • Multiple key value operation
18. Hashes • Maps between string fields and string values • Perfect data type to represent objects • Better strings, effecient memory usage • Useful for objects, resource caching
19. Hashes • Atomic incremention • Listing only keys or values • Operations only by keys of hashes • Multiple hash operations
20. Lists • Simply lists of strings • Sorted by insertion order • Atomic push/pop from head or tail • Useful for queues, array lists
21. Lists • Blocking pop • Trimming • Operations by index • Pop and append to another list
22. Sets • Unordered collection of strings • Not allowing repeated members • Good to represent relations • Useful for friends, favorites, tags
23. Sets • Intersection of sets • Difference of sets • Random element picking • Union of sets
24. Sorted sets • Sorted, unique collections of strings • Each member associated with score • Get ranges by score or by position • Useful for timelines, leader boards
25. Sorted sets • Union of sets • Intersection of sets • Querying by score, range • Removing members by score, rank
26. Pub/Sub • Implements the Publish/Subscribe messaging paradigm • SUBSCRIBE, UNSUBSCRIBE and PUBLISH commands used • Pattern-matching subscriptions supported • Useful for real-time communication
27. Who uses • twitter: Every timeline (800 tweets per uses) is on redis (2TB of RAM) • WoW: 8 servers with 64GB to serve user avatars • Pinterest: Most of the graph data per user and per board (110 instance) • StackOverflow: Three level of cache (local, site, global, 1.2 TB of data)
28. Who uses • Weibo, Github • Engineyard, Disqus • Craigslist, Tumblr • Instagram, hipchat
29. Thanks! http://github.com/o