Local Cache Solutions in Golang: A Comprehensive Guide to Open-Source Components

This article provides an in-depth analysis of local cache solutions for Golang backend development. The author starts by examining the fundamental requirements for local caching in business applications: improving read/write performance, supporting expiration times, implementing eviction policies, and addressing GC (Garbage Collection) concerns in Go.

The article surveys seven major open-source local cache components available for Golang: freecache, bigcache, fastcache, offheap, groupcache, ristretto, and go-cache. A comprehensive comparison table is provided to assist developers in making informed decisions during technology selection.

The core section delves into the implementation principles of four major cache solutions:

Freecache uses 256 segments for data sharding, with each segment maintaining its own mutex lock. It implements a custom pointer-free map using slices to avoid GC overhead, and stores data in a ring buffer with pre-allocated memory. The data flow follows: freecache → segment → (slot, ringbuffer).

Bigcache consists of 2^n cacheShard objects (default 1024), each with an RWLock. It uses map[uint64]uint32 for indexing and stores data in entry ([]byte) ring buffers in TLV format. Unlike freecache, bigcache's ring buffer can auto-expand, though it only supports global expiration windows rather than per-key expiration.

Fastcache is inspired by bigcache and contains 512 buckets, each with its own RWLock. It uses map[uint64]uint64 for indexing and stores data in a 2D slice (chunks) structure. Its key differentiator is heap-free memory allocation using off-heap memory, completely avoiding GC impact. However, it lacks built-in expiration support.

Offheap is a simpler solution that builds a hash table on heap memory allocated via system calls. It uses open addressing with linear probing for collision resolution, resulting in minimal GC overhead.

The article concludes that achieving zero-GC in local cache implementations primarily relies on two approaches: allocating off-heap memory (Mmap) or avoiding GC through map non-pointer optimization (e.g., map[uint64]uint32) or using slice-based map implementations. High performance is achieved through data sharding to reduce lock contention.