Redis is an open source, advanced key-value store. It is often referred to as a data structure server since keys can contain strings, hashes, lists, sets and sorted sets.


Redis is a BSD-licensed, advanced key-value store which works in-memory but provides disk persistence. It is often referred to as a data structure server since keys can contain as values different data structures:

  • Strings, which are binary safe and up to 512 MB in size.

  • Lists, offering O(1) push/pop/trim/length operations regardless of the number of elements contained inside the list. Lists also provide blocking operations (pop-style commands that block if there are no elements in the list), so Redis lists are often used in order to implement background jobs and other kinds of queues. There are very popular libraries like Resque and Sidekiq using Redis as a backend.

  • Hashes are field-value maps like in most programming languages. They are useful in order to represent objects and are very memory efficient for a small number of fields, yet very scalable supporting up to 2.14 billion fields per hash.

  • Sets are unordered collections of elements and are useful in order to add, remove, and test elements for existence in constant-time. Small sets of integers are extremely space efficient, and but sets scale up to 2.14 billion elements per set. It is possible to ask for random elements inside sets which is very useful. See SPOP and SRANDMEMBER for more information.

  • Sorted sets are very useful data structures where collections or elements are ordered by a floating point number called score. The data structure offers a set of very powerful operations running in logarithmic time: it is possible to add and remove elements, increment the score of elements, get ranges by rank and by score, given an element get its position (rank) or score, and so forth. A notable application is leader boards involving million of users: there are companies using Redis sorted sets in order to implement leader boards of popular games such as Facebook games.

  • Counters are not exactly a type per se, but actually operations you can use with strings that represent integers. For example, the command INCR mykey will automatically create a key with the string value "1" if the key does not exist. The next call will modify the value of the string into "2", and so forth. You can increment and decrement by floats or by any amount. Values are in the range of a signed 64-bit number even when using Redis on 32-bit architectures.

  • Bit operations like counters operate in strings in a different way. The user is basically able to treat the string as an array of bits, doing very memory-efficient operations. For example, if you have ten million users and want to store a boolean value for every user, you'll need just a bit more than 1 MB of memory! Because of the rich set of bitwise commands you can: count the number of set bits with BITCOUNT; perform bitwise AND, OR, XOR, and NOT between bitmaps using BITOP; find the first bit clear or set in a given range with BITPOS; and so forth.

  • HyperLogLog is a probabilistic data structure that efficiently (in terms of computational and memory complexity) addresses the count-distinct problem. The Redis implementation of HLL requires only 12KB for each counter and exhibits a standard error of 0.81%. HLLs can be added with items, merged and counted using the PFADD, PFMERGE and PFCOUNT commands, respectively (the PF prefix of the commands is in honor of Phillipe Flajolet, HyperLogLog's inventor).

To get started quickly, try Redis directly inside your browser, read this quick intro to Redis data types, or watch a great presentation by Peter Cooper.

Features as a data store

While Redis is an in-memory system, it offers a lot of features of a data store.

  • Tunable on-disk persistence with a point-in-time snapshotting persistence, or an Append Only File with tunable fsync policy.
  • Asynchronous replication.
  • Redis is also a very fast Pub/Sub server.
  • An API to configure Redis at runtime and automatically rewrite the configuration file.
  • Automatic failover and monitoring via Redis Sentinel.
  • Shared-nothing clustering is available from v3.

It has an impressive ecosystem of client libraries for all the mainstream and elite programming languages.


The Redis community is big and willing to help.


If you are looking for commercial support, please see the community page at for more information