Riak KV stores data as a combination of keys and values, and is a fundamentally content-agnostic database
Riak is a masterless system designed to favor availability, even in the event of node failures and/or network partitions. Any server (“node” in Riak parlance) can serve any incoming request, regardless of data locality, and all data is replicated across multiple nodes. If a node experiences an outage, other nodes will continue to service write and read requests
The Ring State is a data structure that gets communicated and stays in sync between all the nodes, so each node knows the state of the entire cluster. If a node gets a request for an object managed by another node, it consults the Ring State and forwards the request to the proper nodes, effectively proxying the request as the coordinating node.
Each vnode is a separate process which is assigned a partition of the Ring.
"This uniformity of Riak KV vnode responsibility provides the basis for Riak KV’s fault tolerance and scalability. If your cluster has 64 partitions and you are running three nodes, two of your nodes will have 21 vnodes, while the third node holds 22 vnodes"
No single vnode is responsible for more than one replica of an object.
In Riak, data is automatically distributed evenly across nodes using consistent hashing. Consistent hashing ensures data is evenly distributed around the cluster and new nodes can be added with automatic, minimal reshuffling of data. This significantly decreases risky “hot spots” in the database and lowers the operational burden of scaling.
Hashing and shared responsibility for keys across nodes ensures that data in Riak is evenly distributed. When machines are added, data is rebalanced automatically with no downtime. New machines take responsibility for their share of data by assuming ownership of some of the partitions; existing cluster members hand off the relevant partitions and the associated data
Riak KV chooses one vnode to exclusively host a range of keys, and the other vnodes host the remaining non-overlapping ranges. With partitioning, the total capacity can increase by simply adding commodity servers. Since replication improves availability and partitions allow us to increase capacity, Riak KV combines both partitions and replication to work together
Relaxed consistency, also known as eventual consistency, means that not all of the assigned nodes for a transaction in a distributed system need to have that transaction confirmed before the distributed system considers that transaction to be complete. This allows for a higher degree of workload concurrency and data availability
- Riak KV stores data as a combination of keys and values in buckets
- Buckets in Riak KV provide logical namespaces so that identical keys in different buckets will not conflict.
- A unique key in Riak KV is defined by bucket/key.
Riak KV addresses any inconsistencies by returning the most recently updated version, determined by looking at the object’s Dotted Version Vector, or DVV. DVVs are metadata attached to each data replica when it is created. They are extended each time a data replica is updated in order to keep track of data versions.
"A more advanced capability, called Active Anti Entropy, addresses colder data, AAE is a background process which continually compares merkle trees across replica sets to determine discrepancies. Both are key to preventing the need for manual operator intervention under failure scenarios."
To provide high availability, Riak KV defaults to what is known as a sloppy quorum, meaning that if any primary node is unavailable, the next available node in the cluster will accept requests. That node will then update the primary node when it comes back online. This ability to easily handle node failures is known as a hinted handoff.