RS's unique and patented technology was developed to meet these main objectives:

  • Decouple data path from cluster management
  • Ensure consistent top performance
  • Simplify and automate operations

Decouple data path from cluster management

RS's architecture completely decouples the data path from the cluster management in order to optimize the operations of each of these components.

Data path

The data path is based on multiple zero-latency, multi-threaded proxies that reside on each of the cluster nodes to mask the underlying complexity of the system. This model supports multiple proxies per Redis database and permits the use of any regular or cluster-aware Redis client. This allows the user's code base to work as-is. Each proxy forwards client requests to the relevant Redis servers (shards). When new shards are added to a database, the proxy handles the new routing rules for the application's requests transparently, by immediately scaling the database performance and memory capacity without any changes to the application itself.

The proxies also support Memcached's text and binary protocols and perform in-flight translation between Memcached and Redis protocol semantics. This allows Memcached users to enjoy many of the features that are not available with open source Memcached, such as built-in replication, auto-failover, data persistence backups and scaling (out/in) without losing data.

Cluster management

The cluster manager is a sophisticated governing function that provides capabilities such as resharding, rebalancing, auto-failover, rack-awareness, database provisioning, resource management, data persistence configuration, backup and recovery. The cluster manager employs multiple watchdog mechanisms at the cluster node level and at the process level, which guarantees an instant response to events such as node, rack and data center failures, and can handle multiple failover events at the same time. Because the cluster manager is entirely decoupled from the data path components, changes to its software components do not affect the data path components.

Consistent top performance

Multiple techniques are implemented by RS to provide consistent top performance for Redis:

  • A shared-nothing architecture maximizes the performance of each database. Automatic migration of shards between nodes is performed when needed.
  • The zero-latency proxy utilizes several mechanisms for improving performance, including just-in-time pipelining, socket connections, connection pooling, and multiplexing.
  • Improvements to AOF data persistence and optimized rewrite algorithms.
  • Diskless replication is employed between master and slave shards, instead of using the default file-based data streaming.
  • File-system-level improvements enable optimal access to storage and support burst write operations without degrading (or blocking) database performance.

Simplified and automated operations

RS boosts the productivity of developers and DevOps by simplifying and automating certain complex and time-consuming ops-related tasks, including provisioning new databases, updating database configuration, resharding existing databases and rebalancing shards across cluster nodes.

The homogeneous, symmetric cluster architecture of RS enables the underlying infrastructure to be fully heterogeneous, where each node in the cluster can be physical or virtual, with variations in RAM capacity, storage devices, and the number of compute cores. This freedom of configuration allows the operator to roll out clusters that make the most efficient use of the resources, with minimal system limitations. For example, a cluster may be based on a few permanent bare-metal servers with additional virtual servers for scaling out (or in) when load increases or decreases. Built-in configurable monitoring functions help the administrator keep an eye on the resource utilization levels in the cluster.

RS also makes sure that there is uninterrupted operation of Redis databases by protecting against out-of-memory events and tuning memory fragmentation automatically.

All of the above is complemented by an easy-to-use web-based user interface, CLI tools, and APIs. Special experience or knowledge is not required to create, manage and monitor Redis clusters, nodes, and databases in RS.

You can read more about the architecture of RS in Concepts and Architecture.