Radix — drop-in Redis replacement on a shard-per-core database

Picodata Radix speaks the Redis wire protocol. Drop-in for Redis Cluster: clients keep their code; the database takes over scale, durability, and operations.

Download Picodata Radix documentation

Key Radix advantages

  • Drop-in Redis replacement

    Radix speaks RESP2 and the Redis Cluster client protocol. Clients connect with redis-cli, redis-benchmark, and the standard client libraries — no code changes required.

  • Shard-per-core throughput

    Underneath, each Picodata core serves its own independent shard with no contention on shared resources. Linear horizontal scaling on commodity hardware — well past the single-thread ceiling that bounds open-source Redis.

  • Strong consistency option

    Picodata's Raft-replicated WAL gives Radix high availability and zero-data-loss guarantees on top of the Redis API. Run as a cache or as durable storage.

  • Cluster operations

    Add and remove cores or nodes online; the cluster rebalances buckets without downtime. No separate Sentinel infrastructure — failover is built in.

  • Cache or persistent store

    Configure Radix as an in-memory cache or as a persistent key-value store with the same plugin and the same API. Same wire protocol, two operational profiles.

Radix on Picodata

Picodata is an open-source distributed SQL database with PostgreSQL wire compatibility, shard-per-core architecture, and a Rust plugin SDK.

Radix is a commercial plugin from the Picodata team. It proves the model — a wire-compatible Redis layer that runs against the same shared cluster as the SQL core. The SDK is open; the community can build its own wire-compatibility plugins.

What Radix is

Radix is a plugin for Picodata that implements the Redis wire protocol (RESP2) and the Redis Cluster client model. Connect with popular client libraries — redis-cli, ioredis, lettuce, go-redis, and others; the Picodata cluster behaves like Redis Cluster, supporting more than 180 commands and the core data types.

What Radix changes underneath:

  • Automatic horizontal scaling on the Redis Cluster sharding model. Linear throughput growth as cores or nodes are added — verified to operate stably at clusters up to 10 000 cores.
  • Logs operations for observability, audit, and compliance.
  • High availability and fault tolerance built into Picodata — write-ahead log, Raft consensus, replication. No separate Sentinel infrastructure or third-party watchdog needed.
Why Radix is more than Redis

Radix gives you Redis commands and semantics on a database with cluster-wide guarantees. The same security perimeter, role model, audit trails, and consistency contract apply to every key your Redis client touches. Radix is not just wire-compatible — it's a full component of a distributed SQL database.

Where it fits

Radix is a fit wherever the Redis API is required alongside cluster guarantees and operational maturity:

  • Hot-data caches in front of slower databases — with WAL durability, not best-effort eviction;
  • Session stores and rate-limiting counters in high-throughput web applications;
  • Replacement for commercial Redis Enterprise licenses;
  • Consolidating a Redis layer into the same operational footprint as a PostgreSQL-compatible distributed SQL — one cluster, one team.
Who it is for

Radix is aimed at architects, SRE teams, and platform engineers in financial services, telecom, gaming, ad-tech, and consumer-internet — wherever you need Redis on the wire, with HA, WAL, and audit on top.

Architecture and scaling

Radix runs as a plugin inside Picodata cluster nodes. Each node serves its share of keys on a Redis-Cluster-compatible scheme; clients see the same sharding and MOVED redirects. The difference is underneath: each CPU core serves its own shard, with nothing shared. That breaks the Redis single-thread ceiling and lets throughput scale linearly with the cluster.

When Radix is not the right choice

Radix is not a fit if:

  • Only a narrow Redis feature set is needed and a single open-source Redis process is enough;
  • The application code relies on Redis modules (RedisJSON, RediSearch, RedisGears) — Radix does not implement them yet;
  • The workload fits on a single server and there are no fault-tolerance, audit, or scale requirements — at that size the operational investment in Picodata is excessive.

Outside those cases, Radix earns its keep — both technically and on cost.

Quick start

The Picodata documentation site has step-by-step instructions for running Picodata with plugin support and installing and activating Radix. Once installed, connect to Picodata exactly as you would to a regular Redis instance:

redis-cli -p 7379

Core Redis commands are supported: GET, SET, DEL, EXPIRE, HGET, HSET, TTL, SCAN, TYPE, and others.

System requirements

Linux distributions on x86-64. The full list of supported operating systems is on the download page.

What's new in Radix 1.0

Radix 1.0 is our first LTS release with extended lifecycle. The new version supports TLS/mTLS connections, advanced security for users and commands via ACL, eviction policies for keys, as well as better performance coupled with lower memory usage. See Radix documentation for more details.

Feedback and support

Product questions, pilot requests, and bug reports — open an issue on the Picodata GitLab or write to hello@picodata.io.