Redis vs Memcached: Which In‑Memory Database Wins for Your Projects?

Introduction

Both Redis and Memcached are in‑memory databases known for extremely fast access, but developers often wonder which one to choose for a given scenario.

Why is Redis more popular today? This article compares the two from several angles to help you decide.

Thread Model

Memcached uses a multithreaded model with I/O multiplexing; a main thread distributes requests to worker threads, allowing long‑running requests not to block others, though thread switching and locking add overhead.

Redis also uses I/O multiplexing but processes requests in a single thread, meaning a long request blocks the server. The single‑threaded design avoids context‑switch costs and lock contention but cannot exploit multiple CPU cores.

Because Redis is memory‑bound, its bottleneck is memory and network bandwidth rather than CPU.

Data Structures

Memcached only supports simple string values (max 1 MB) with expiration up to 30 days, requiring manual serialization for complex data.

Redis offers a rich set of native types: string, list, hash, set, sorted set, geo, and HyperLogLog, each with dedicated operations, providing great flexibility.

Eviction Policies

Memcached enforces a global memory limit and evicts using LRU, but its design can evict newly written items first.

Redis has no mandatory memory cap and provides multiple eviction strategies such as volatile‑lru, allkeys‑lru, volatile‑random, allkeys‑random, volatile‑ttl, volatile‑lfu, and allkeys‑lfu, allowing selection based on workload.

Pipelining and Transactions

Redis supports pipelining—batching multiple commands in a single network round‑trip—and a lightweight transaction model that ensures atomic execution of a command sequence.

Persistence

Memcached does not persist data; a crash loses everything.

Redis offers RDB snapshots and AOF append‑only logs, which can be combined to guarantee data durability.

High Availability

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

Redis provides master‑slave replication, Sentinel for automatic failover, and read‑write splitting to improve availability and performance.

Clustering

Both systems can run in clusters. Memcached relies on client‑side consistent hashing; Redis uses hash slots distributed across nodes, with each node typically having a replica for added resilience. Adding or removing nodes requires manual rebalancing.

Conclusion

Redis delivers a richer feature set and comparable performance, making it the preferred choice for applications needing diverse data structures and high availability, while Memcached remains suitable for simple, high‑throughput get/set workloads with modest memory requirements.