Redis vs Memcached: A Comprehensive Comparison

Introduction

Redis and Memcached are both in‑memory databases known for extremely fast access speeds, but choosing between them can be confusing. This article compares them from multiple angles to help you select the best fit for your workload.

Comparison Dimensions

Thread model

Data structures

Eviction policy

Pipelining and transactions

Persistence

High availability

Clustering

Thread Model

Memcached uses a multithreaded model with I/O multiplexing; the main thread accepts requests and distributes them to worker threads, which avoids blocking but incurs context‑switch overhead and lock contention.

Redis also uses I/O multiplexing but processes requests in a single thread, eliminating context switches and lock contention, though a long‑running command blocks the whole server. Redis 6.0 adds optional multithreading for network I/O, improving performance.

Data Structures

Memcached only supports simple string values (max 1 MB) and requires manual serialization, offering a "store‑as‑is" model.

Redis provides a rich set of native data types: string, list, hash, set, sorted set, geo, and HyperLogLog, enabling flexible operations without extra serialization.

Eviction Policies

Memcached enforces a global memory limit and uses an LRU algorithm, which can evict recently written keys due to its memory‑management design.

Redis does not require a fixed memory cap and offers multiple eviction strategies, including volatile‑lru, allkeys‑lru, volatile‑random, allkeys‑random, volatile‑ttl, volatile‑lfu, and allkeys‑lfu, allowing you to tailor eviction to your workload.

Pipelining and Transactions

Redis supports pipelining, allowing a client to batch multiple commands in a single network round‑trip, reducing latency.

Redis also provides a lightweight transaction model (different from strict RDBMS transactions) that works together with pipelining to ensure ordered execution and abort on conflicting key modifications.

Persistence

Memcached does not persist data; a crash results in total data loss.

Redis offers two persistence mechanisms: RDB snapshots (full dumps) and AOF (append‑only file) for incremental logging, which can be combined to maximize data safety.

High Availability

Memcached lacks built‑in replication; it must be deployed as a single node or with custom client‑side sharding.

Redis provides master‑slave replication, Sentinel for automatic failover, and optional read‑write splitting, greatly improving availability.

Clustering

Both systems can run in clusters, but the approaches differ.

Memcached relies on client‑side consistent hashing; when a node fails, other nodes absorb its traffic.

Redis Cluster distributes hash slots across nodes, each node having at least one replica, and requires manual resharding when nodes are added or removed. Third‑party proxy solutions like Codis or Twemproxy also exist.

Summary Table

#

Memcached

Redis

Thread model

Multithreaded

Single‑threaded (optional multithreaded I/O in Redis 6.0)

Data structures

String only, max 1 MB, 30‑day TTL limit

String, list, hash, set, sorted set, geo, HyperLogLog

Eviction policy

LRU

LRU, LFU, random, TTL‑based, etc.

Pipelining & transactions

Not supported

Supported

Persistence

Not supported

RDB & AOF

High availability

Not supported

Master‑slave replication + Sentinel

Clustering

Client‑side consistent hashing

Redis Cluster (hash slots) + replicas

Conclusion

Redis offers a richer feature set with comparable performance to Memcached, making it the preferred choice for most modern applications that need diverse data structures, persistence, and high availability. Memcached remains suitable for simple, high‑throughput get/set workloads with minimal memory usage.