Based on the Remote Dictionary Server technology, Scaleway Database for Redis® stores your data in the RAM of the underlying machine rather than on a disk (SSD/HDD). In other words, for each request to read, to insert, or to update data in a database, this can be executed using data available in the fastest and closest storage of your compute resource, the memory.

Traditional databases like MySQL or PostgreSQL store data on a disk which inevitably introduces IOPs and results in latency on each operation. Redis® is known for delivering millisecond response time and high performance for millions of requests per second to empower demanding workloads. The combination of a powerful in-memory data storage such as Redis and the management of the resources set-up, securisation, scaling and maintenance makes Scaleway Database for Redis a handy solution to improve the usability of your application.

One of the most common ways to implement cache is storing frequently accessed data in Redis® (therefore in memory), and serving your application's request from there, should data be available, instead of from your traditional database. If data is not available in memory, it can always be retrieved from the primary database.

A Redis™ cluster contains a minimum of 3 nodes and up to 6 nodes. Each node contains a source and a replica. The cluster nodes use hash partitioning to split the keyspace into key slots. Each replica copies the data of a specific source and can be reassigned to replicate another source or be elected as a source node if needed. This is much better for scaling as the operation is spread across multiple nodes instead of having a single entry point.

Two-node High Availability configurations are available with Redis™ Database Instances. This configuration type allows you to create a standby node, with an up-to-date replica of the database. If the main node fails for any reason, the standby can take over requests, reducing downtime.

The HA standby node is linked to the main node, using asynchronous replication. Asynchronous replication allows you to maintain good performance.

A Redis® cluster contains a minimum of 3 nodes and up to 6 nodes. Each node contains a source and a replica. The cluster nodes use hash partitioning to split the keyspace into 16,384 key slots. Each source of the cluster is responsible for a subset of those slots. Each replica copies the data of one of the sources. For example, on a three-node Redis® Database Instance cluster, three Instances host each a source and a replica of one of the other nodes’ sources. If one of the sources fails, the remaining nodes hold a vote and elect the replica that will be promoted as the new source. When the failing source rejoins the cluster, it automatically becomes a replica. It begins to copy the data of the source of another node.

You can scale your cluster horizontally up to six nodes. Below is an example of a configuration for a three nodes cluster:

Instance A contains hash slots from 0 to 5,500
Instance B contains hash slots from 5,501 to 11,000
Instance C contains hash slots from 11,001 to 16,383
Each of the three Instances acts as a primary node and replicates one of the others as a secondary node.