Below is a list of terms that you may run into frequently in the documentation for Riak, along with links to more in-depth treatments.
Active Anti-Entropy (AAE)
A continuous background process that compares and repairs any divergent, missing, or corrupted replicas. Unlike read repair, which is only triggered when data is read, the Active Anti-Entropy system ensures the integrity of all data stored in Riak. This is particularly useful in clusters containing “cold data,” i.e. data that may not be read for long periods of time, potentially years. Furthermore, unlike the repair command, Active Anti-Entropy is an automatic process requiring no user intervention. It is enabled by default in Riak 1.3 and greater.
Basho Bench is a benchmarking tool created to conduct accurate and repeatable performance tests and stress tests and to produce performance graphs.
A bucket is a namespace for data stored in Riak, with a set of common
properties for its contents, e.g. the number of replicas (
whether siblings are returned on reads (
allow_mult), etc. Buckets'
properties are determined by their bucket type (see below).
Bucket types enable you to create and manage sets of bucket properties that, when applied to buckets, dictate those buckets' behavior. They also act as a third namespace in Riak in addition to buckets and keys.
A Riak cluster is a 160-bit integer space which is divided into equally-sized partitions. Each vnode in the Riak Ring is responsible for one of these partitions.
Consistent hashing is a technique used to limit the reshuffling of keys when a hash-table data structure is rebalanced (i.e. when slots are added or removed). Riak uses consistent hashing to organize its data storage and replication. Specifically, the vnodes in the Riak Ring responsible for storing each object are determined using the consistent hashing technique.
Riak Data Types are data objects inspired by research on CRDTs that use certain rules of convergence to dictate how conflicts between replicas are resolved in Riak's eventually consistent system. There are five Riak Data Types in total: flags, registers, counters, sets, and maps.
A consistency model that informally guarantees that if no new updates are made to a given data item, all reads on that item will eventually return the last updated value. Details about what this means in Riak can be found in the document below.
Riak uses a “gossip protocol” to share and communicate ring state and bucket properties around the cluster. Whenever a node changes its claim on the ring, it announces its change via this protocol. Each node also periodically sends its current view of the ring state to a randomly selected peer in case any nodes missed previous updates.
Hinted handoff is a technique for dealing with node failure in the Riak cluster in which neighboring nodes temporarily take over storage operations for the failed node. When the failed node returns to the cluster, the updates received by the neighboring nodes are handed off to it.
Hinted handoff allows Riak to ensure database availability. When a node fails, Riak can continue to handle requests as if the node were still there.
Keys are unique object identifiers in Riak and are scoped within buckets and bucket types.
Lager is an Erlang/OTP framework that ships as Riak's default logger.
Riak's MapReduce gives developers the capability to perform more powerful queries over the data stored in their key/value data.
A node is analogous to a physical server. Nodes run a certain number of nodes, each of which claims a partition in the Riak Ring key space.
An object is another name for a value.
Partitions are the spaces into which a Riak cluster is divided. Each
vnode in Riak is responsible for a partition. Data is stored on a set
number of partitions determined by the
n_val setting, with the target
partitions chosen statically by applying consistent hashing to an
Quorum in Riak has two meanings:
- The quantity of replicas that must respond to a read or write request before it is considered successful. This is defined as a bucket property or as one of the relevant parameters to a single request (R,W,DW,RW).
A symbolic quantity for the above,
quorum, which is equivalent to
n_val/ 2 + 1. The default setting is
During failure scenarios, in which available nodes < total nodes, sloppy quorum is used to ensure that Riak is still available to take writes. When a primary node is unavailable, another node will accept its write requests. When the node returns, data is transferred to the primary node via the Hinted Handoff process.
Read repair is an anti-entropy mechanism that Riak uses to optimistically update stale replicas when they reply to a read request with stale data.
Replicas are copies of data stored in Riak. The number of replicas required for both successful reads and writes is configurable in Riak and should be set based on your application's consistency and availability requirements.
Riak Core is the modular distributed systems framework that serves as the foundation for Riak's scalable architecture.
Riak KV is the key/value datastore for Riak.
Riak Pipe is the processing layer that powers Riak's MapReduce. It's best described as “UNIX pipes for Riak.”
Riak Search is a distributed, scalable, failure-tolerant, realtime, full-text search engine integrating Apache Solr with Riak KV.
The Riak Ring is a 160-bit integer space. This space is equally divided into partitions, each of which is claimed by a vnode, which themselves reside on actual physical server nodes.
Secondary Indexing (2i)
Secondary Indexing in Riak gives developers the ability to tag an object stored in Riak with one or more values which can then be queried.
While Riak is most well known as an eventually consistent data storage system, versions of Riak 2.0 and greater enable you to apply strong consistency guarantees to some or all of your data, thus using Riak as a CP (consistent plus partition-tolerant) rather than AP (highly available plus partition-tolerant) system.
Riak is best described as a key/value store. In versions of Riak prior to 2.0, all “values” are opaque BLOBs (binary large objects) identified with a unique key. Values can be any type of data, including a string, a JSON object, a text document, etc. Modifying values involves fetching the value that exists in Riak and substituting it for a new value; operations on values are thus basic CRUD operations.
Riak Data Types, added in version 2.0, are an important exception to this. While still considered values—because they are stored in bucket type/bucket/key locations, like anything in Riak—Riak Data Types are not BLOBs and are modified by Data Type-specific operations.
Riak utilizes vector clocks (or vclocks) to handle version control. Since any node in a Riak cluster is able to handle a request, and not all nodes need to participate, data versioning is required to keep track of a current value. When a value is stored in Riak, it is tagged with a vector clock and establishes the initial version. When it is updated, the client provides the vector clock of the object being modified so that this vector clock can be extended to reflect the update. Riak can then compare vector clocks on different versions of the object and determine certain attributes of the data.
Vnodes, or “virtual nodes,” are responsible for claiming a partition in the Riak Ring, and they coordinate requests for these partitions. Vnodes reside on physical nodes in a Riak cluster, and the number of vnodes per physical node is determined by the total number of vnodes and the number of active physical nodes in the cluster. Riak balances the assignment of vnodes across the active physical nodes.