Distributed Systems active

RateShield — High-Performance Distributed Rate Limiter

An ultra-fast, zero-allocation Go rate limiter achieving sub-40ns execution time using a 64-shard in-memory engine, atomic Redis Lua scripting, and multi-tier failover.

Key Architectural Highlights

  • ✓ Sub-40ns execution time (34.84ns Token Bucket)
  • ✓ 0 B/op allocations on hot path
  • ✓ 64-shard FNV-1a mutex engine
  • ✓ Atomic single-RTT Redis Lua scripts
  • ✓ Multi-tier resilient fallback
GoRedisLuaFNV-1a Hashingnet/httpGinFiberEchoChi
View GitHub Repository ↗
BENCHMARK TOOL

Throughput & Allocation Analyzer

Intel Core i5-12450H Data

Simulate traffic scale to evaluate latency, CPU cycle consumption, and heap allocation pressure across rate limiting algorithms.

Execution Latency 34.84 ns Per evaluation decision
Heap Memory Allocated 0 B/sec Zero GC pressure
CPU Time Spent Limiting 3.48 ms/sec <0.35% of a single CPU core

Overview

RateShield is an open-source, production-grade distributed rate limiting library for Go. It is architected from the ground up for high-throughput, low-latency API gateways and microservices that cannot tolerate garbage collection pauses or lock contention under heavy concurrency.

Architecture

RateShield utilizes the Strategy Pattern combined with Dependency Inversion, decoupling rate limiting algorithms (Token Bucket, Fixed Window, Sliding Window Counter, Sliding Window Log) from backend storage implementations:

 ┌──────────────────────────────────────────────────────────────────┐
 │                        HTTP Request / API                        │
 └────────────────────────────────┬─────────────────────────────────┘
                                  │
                                  ▼
 ┌──────────────────────────────────────────────────────────────────┐
 │                   Middleware / Key Extractor                     │
 │          (net/http, Gin, Fiber, Echo, Chi | IP / User)           │
 └────────────────────────────────┬─────────────────────────────────┘
                                  │
                                  ▼
 ┌──────────────────────────────────────────────────────────────────┐
 │                        rateshield.Limiter                        │
 └────────────────────────────────┬─────────────────────────────────┘
                                  │
         ┌────────────────────────┼────────────────────────┐
         ▼                        ▼                        ▼
 ┌───────────────┐        ┌───────────────┐        ┌───────────────┐
 │ Token Bucket  │        │ Fixed Window  │        │Sliding Window │
 └───────┬───────┘        └───────┬───────┘        └───────┬───────┘
         │                        │                        │
         └────────────────────────┼────────────────────────┘
                                  │
                                  ▼
 ┌──────────────────────────────────────────────────────────────────┐
 │                         store.Store                              │
 │   (Sharded Memory | Atomic Redis Lua | Resilient Fallback)       │
 └──────────────────────────────────────────────────────────────────┘

Key Technical Innovations

  1. 64-Shard Mutex Memory Engine: Partitions the hash space into 64 distinct lock domains using 64-bit FNV-1a hashing, eliminating cross-core lock contention.
  2. Zero-Allocation Execution (0 B/op): Exploits value receivers and stack allocation semantics to ensure zero heap pressure during rate evaluations.
  3. Atomic Distributed Redis Lua: Single round-trip evaluation guarantees race-condition-free operation across distributed Kubernetes replicas.
  4. Resilient Failover: Automatically degrades from Redis cluster to localized sharded memory if network partitions occur.
  5. Modern IETF Headers: Supports modern RateLimit-Limit, RateLimit-Remaining, RateLimit-Reset standard headers alongside legacy X-RateLimit-*.

Benchmark Details

Evaluated on 12th Gen Intel(R) Core(TM) i5-12450H (go test -bench=. -benchmem ./algorithms):

  • Token Bucket: 34.84 ns/op | 0 B/op | 0 allocs
  • Fixed Window: 35.61 ns/op | 0 B/op | 0 allocs
  • Sliding Window Counter: 38.16 ns/op | 0 B/op | 0 allocs
  • Sliding Window Log: 175.70 ns/op | 48 B/op | 1 alloc