redis-caching.md

Redis Caching — Performance Comparison

Milestone M3.5 | Analyst: Backend Lead | Date: May 2026

Implementation

Cache-aside pattern via RedisCacheService in Loopless.Infrastructure/Caching/RedisCacheService.cs.

Request → Handler → Check Redis → [HIT] return cached response
                               → [MISS] query PostgreSQL → store in Redis → return response

Endpoint benchmarked: GET /api/v1/users/me (user profile fetch)

Cache key: user:profile:{userId}

TTL: 15 minutes (defined as UserProfileTtl in RedisCacheService)

Benchmark Setup

• Tool: k6 (script: devops/loadtest/k6-script.js)
• Scenario: 50 concurrent VUs, 60 seconds sustained
• Target: authenticated requests to /api/v1/users/me
• Environment: local Docker Compose (docker compose up)

Results

Without Redis (cache disabled — direct PostgreSQL each request)

http_req_duration (p50=42.1ms, p90=87.3ms, p95=112.4ms, p99=198.7ms)
http_reqs............: 12,843  total (214 req/s avg)
checks...............: 100% ✓ 12843 ✗ 0

With Redis (cache-aside, 15 min TTL)

http_req_duration (p50=2.3ms, p90=4.1ms, p95=5.8ms, p99=11.2ms)
http_reqs............: 13,921  total (232 req/s avg)
checks...............: 100% ✓ 13921 ✗ 0

Cache Miss Behavior

First request after TTL expiry: ~45ms (DB read + cache write overhead ≈ 3ms) Subsequent requests: ~2.3ms (cache hit)

Cache invalidation: RedisCacheService.RemoveAsync($"user:profile:{userId}") called on profile update.

Summary

Metric	No Cache	With Redis	Improvement
p50 latency	42.1ms	2.3ms	94.5%
p95 latency	112.4ms	5.8ms	94.8%
p99 latency	198.7ms	11.2ms	94.4%
Throughput	214 req/s	232 req/s	+8.4%

Other Cached Endpoints

Endpoint	Cache Key	TTL
GET /api/v1/users/me	user:profile:{userId}	15 min
GET /api/v1/skills	skills:all	5 min
GET /api/v1/matching/discover	matching:discover:{userId}:{page}	5 min