Cache stampede prevention in HLD - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the primary cause of a cache stampede?

Imagine many users request the same data at the same time, but the cache for that data has expired. What causes a cache stampede in this scenario?

A The database is down and cannot respond to any requests.

B The cache server crashes due to overload from unrelated requests.

C Multiple requests simultaneously miss the cache and all try to fetch data from the database at once.

D The cache data is corrupted and cannot be read by any request.

💡 Hint

Think about what happens when many requests try to get data that is not in the cache.

A cache stampede happens when many requests miss the cache at the same time and all try to get the data from the database, causing a sudden load spike.

❓ Architecture

intermediate

Which design best prevents cache stampede by allowing only one request to refresh the cache?

Choose the architecture pattern that ensures only one request fetches fresh data from the database while others wait for the cache to be updated.

A Disable caching entirely to avoid stale data.

B Use a distributed lock or mutex around cache refresh operations.

C Allow all requests to fetch data from the database simultaneously.

D Use a random delay before each request tries to refresh the cache.

💡 Hint

Think about how to coordinate requests so only one updates the cache at a time.

Using a distributed lock ensures only one request refreshes the cache, preventing multiple database hits.

❓ scaling

advanced

How does 'early recomputation' help in cache stampede prevention?

Consider a cache system that refreshes data before it expires. What is the main benefit of this approach?

A It reduces the chance that many requests find the cache expired simultaneously.

B It increases the cache expiration time to avoid frequent refreshes.

C It disables caching for high-traffic data to prevent stale reads.

D It forces all requests to bypass the cache and go to the database.

💡 Hint

Think about how refreshing cache early affects request timing.

Early recomputation refreshes cache before expiration, so requests rarely see expired cache and avoid stampedes.

❓ tradeoff

advanced

What is a tradeoff when using 'request coalescing' to prevent cache stampede?

Request coalescing groups multiple requests for the same data to a single database fetch. What is a downside of this approach?

A It can increase latency for some requests waiting for the first to complete.

B It causes the cache to expire more quickly than normal.

C It requires disabling the cache for all requests.

D It duplicates database queries for each request.

💡 Hint

Consider what happens to requests that wait for others to finish fetching data.

Request coalescing reduces load but can increase wait time for some requests, causing higher latency.

❓ estimation

expert

Estimate the maximum database load during a cache stampede for 10,000 simultaneous requests with no prevention.

If 10,000 users request the same expired cache key at the same time, and each cache miss triggers one database query, what is the maximum number of database queries generated?

A 1 query

B 1,000 queries

C 100 queries

D 10,000 queries

💡 Hint

Think about how many requests miss the cache and query the database simultaneously.

Without prevention, all 10,000 requests miss the cache and each triggers a database query, causing 10,000 queries.