Перейти до основного вмісту

Caching Strategy: Redis & App-level

This document outlines the caching layers and patterns to optimize performance while ensuring data consistency.

1. Cache Layers​

  1. Redis Read Cache

    • Stores seat availability for quick CheckAvailability responses.
    • TTL: 10 seconds to limit staleness.
    • Key: seat:{seat_id} → { "available": true/false }.

  2. Application-Level Cache

    • Local in‑memory cache (e.g., Caffeine, Guava) for hot routes.
    • TTL: 1 minute; Max Size: 1000 entries.

2. Caching Patterns​

2.1. Cache-Aside Pattern​

Application:
  on request:
    value = cache.get(key)
    if value == null:
      value = database.read(key)
      cache.put(key, value, TTL)
    return value

2.2. Cache Stampede Protection​

  • Mutex Lock: Use Redis SETNX to lock on cache miss.
  • Request Coalescing: Use fallback TTL for stale responses.

3. Consistency & Invalidation​

  • Eviction: LRU policy with TTL and size limits.
  • Invalidation: On seat lock/unlock, delete seat:{seat_id} key.
  • Write-Through: Cache updated after DB transaction commit.
  • Document Version: 1.0
  • Date: 2025-06-23
  • Author: ArturChernets