Choosing Between Memcached and Redis: Which Cache Wins for Massive Datasets?

Problem

Database tables contain massive data (tens of millions of rows) and require faster server response to user requests.

Solution

Use high‑speed server cache to cache database data.

In‑memory database.

(Only considering data caching; later can adopt Hadoop+HBase+Hive and other distributed storage analysis platforms.)

Comparison of Mainstream Cache and Database Solutions

The technologies shown represent current data storage implementations, including relational databases (MySQL/PostgreSQL), NoSQL databases (MongoDB), in‑memory databases (Redis), and in‑memory caches (Memcached). For large tables we need high‑efficiency query speed; ordinary relational databases cannot satisfy this. MongoDB, while capable of storing massive data, is not suitable for caching scenarios.

From the above data, two feasible technical solutions for our product emerge:

Memcached – in‑memory key‑value cache

Redis – in‑memory database

Detailed Analysis of Memcached and Redis

Memcached Overview

Memcached is a high‑performance distributed memory object caching system used by dynamic web applications to reduce database load. It caches data and objects in memory, decreasing database reads and accelerating site performance. It is employed by companies such as LiveJournal, Facebook, Vox, and others.

How Memcached Works

Many web applications store data in an RDBMS; application servers read the data and render it to browsers. As data volume and access concentration increase, the RDBMS becomes a bottleneck, leading to slower responses. Memcached caches database query results, reducing the number of database accesses, thereby improving speed and scalability. The diagram below illustrates the collaboration between Memcached and the database.

Process:

Check whether the requested data exists in the cache; if it does, return it directly without querying the database.

If the data is not in the cache, query the database, return the result, and store a copy in the cache.

Maintain cache freshness by updating or invalidating cached entries whenever the underlying data changes.

Key characteristics of Memcached:

Simple protocol

Event handling based on libevent

Built‑in memory storage

Distributed without inter‑node communication

Scalability of Memcached

Distribution is implemented on the client side via an algorithm that maps keys to target nodes.

Redis Overview

Redis is a key‑value storage system. Like Memcached, it stores data in memory for speed, but it supports a richer set of value types, including string, list, set, and sorted set. These types support atomic push/pop, add/remove, and set operations, as well as sorting. Redis periodically persists data to disk or an append‑only log and provides master‑slave replication. It is widely used by companies such as Sina, Taobao, Flickr, and GitHub.

How Redis Works

Redis features:

Multiple data models

Persistence (RDB snapshots or AOF log)

Master‑slave synchronization

Redis supports five primary data types: String, Hash, List, Set, and Sorted Set. Data is usually stored in memory, though virtual memory can be used. Persistence can be achieved via snapshots (high performance, possible data loss) or AOF logs (safer, slower). Redis can synchronize data to multiple replica nodes, improving read performance.

Redis Scalability

Versions prior to 2.8 used client‑side distribution similar to Memcached, optionally with proxies such as Twemproxy. Since version 3.0, Redis Cluster provides server‑side sharding without a single point of failure. The cluster divides the key space into 16,384 hash slots, allowing up to 16,384 nodes.

Conclusion

Both Memcached and Redis can effectively address the performance problem. Redis can be viewed as an extended, more feature‑rich version of Memcached. Specific comparisons:

Performance: Redis performs better for small values on a single core, while Memcached outperforms Redis for payloads larger than 100 KB.

Memory & Data Size: Memcached uses LRU and adjustable max memory; Redis adds virtual memory, breaking physical limits.

Ease of Use: Memcached stores only simple cache data; Redis offers richer data structures, reducing network I/O and enabling in‑server data manipulation.

Reliability: Memcached lacks persistence (data lost on power loss); Redis provides persistence and recovery, at some performance cost.

Application Scenarios: Memcached is ideal for read‑heavy, write‑light caching of large datasets (e.g., user profile queries). Redis suits workloads requiring high read/write efficiency, complex data processing, and stronger data safety (e.g., counting and posting services).

Considerations

Memcached limits a single value to 1 MB, whereas Redis supports up to 512 MB per value.

Memcached is a pure cache with no reliability guarantees; Redis, as an in‑memory database, demands higher reliability.

Fundamentally, Memcached is a simple key‑value cache; Redis is a data‑structure in‑memory database supporting multiple types and operations.

Redis 3.0+ cluster balances client requests across nodes, enabling scalable and maintainable deployments.