Q8: Cache Stampede, Thundering Herd, and Cache Invalidation Strategies. - System Design (50 Qs)

In Simple Words:

Cache Stampede (Thundering Herd): When a heavily requested hot cache key expires (e.g. homepage top stories or flash sale product), thousands of concurrent requests experience a cache miss simultaneously and hammer the primary database, crashing it.
Cache Penetration: Malicious or buggy clients query keys that exist neither in cache nor in the database (e.g. id = -9999), constantly hitting the DB.
- Fix: Cache null objects with short TTL (e.g. 60s) or filter queries using Bloom Filters.
Cache Breakdown: A single super-hot key expires.
- Fix: Mutex distributed lock or Probabilistic Early Recomputation (XFetch).

typescript

// Thundering Herd Protection using Mutex Distributed Lock in Node.js
export async function getWithMutexLock<T>(
  key: string,
  dbFetcher: () => Promise<T>,
  redis: any,
  ttl: number
): Promise<T> {
  let val = await redis.get(key);
  if (val) return JSON.parse(val);

  const lockKey = `lock:${key}`;
  // Try acquiring lock for 5 seconds
  const acquired = await redis.set(lockKey, 'locked', 'NX', 'PX', 5000);

  if (acquired) {
    try {
      const data = await dbFetcher();
      await redis.set(key, JSON.stringify(data), 'EX', ttl);
      return data;
    } finally {
      await redis.del(lockKey);
    }
  } else {
    // Another worker is fetching from DB: sleep and retry
    await new Promise(resolve => setTimeout(resolve, 80));
    return getWithMutexLock(key, dbFetcher, redis, ttl);
  }
}