Caching

Cards

1. Walk through a read with the cache-aside pattern. easy Flashcard
2. What do you trade when you choose a cache TTL? easy Flashcard
3. Why do sites name static files with a content hash, like app.3f9a2c.js? easy Flashcard
4. A service gets 100,000 reads/s, and its cache hit rate is 95%. How many reads per second reach the database? easy Estimate
5. Storing "this key does not exist" in the cache, so repeated lookups for a missing item do not all reach the database, is called _____. easy Fill in the blank
6. A cache hit takes 1 ms. A miss takes 20 ms (cache check plus database read). With a 90% hit rate, what is the average read latency? medium Estimate
7. Which decision should never be made from a cache alone? medium Multiple choice
8. When a hot key expires and thousands of requests miss at once and all hit the database, it is a _____. Letting one request rebuild the value while the others wait for it is called _____. medium Fill in the blank
9. You want CDNs to cache a page for a day but browsers for only a minute. Which header does that? medium Multiple choice
10. Why can restarting a whole cache cluster take down the database behind it, and how do you avoid that? medium Flashcard
11. A feed cache stores each user's feed as a list of post ids, and the posts themselves in a separate cache. Why split them? medium Multiple choice
12. When would you choose Redis over Memcached as a cache? medium Flashcard
13. You add an in-process cache on each app server in front of a shared Redis. What do you gain and what do you pay? medium Flashcard
14. A job imports 50 million records that are rarely read in the days after. Which cache write policy fits? medium Multiple choice
15. Which cache write policy can lose acknowledged writes if the cache node crashes? medium Multiple choice
16. After updating a row in the database, why do most cache-aside systems delete the cache key rather than write the new value into it? hard Multiple choice
17. To keep a hot key from expiring under load, a reader can refresh it at random shortly before its TTL ends, more likely the closer expiry is. This is _____. hard Fill in the blank
18. One cache key gets 1 million reads a second, more than a single cache node can serve. What are the usual fixes? hard Flashcard
19. What is the weakness of plain LFU eviction, and how do real caches handle it? hard Flashcard
20. Every night a batch job reads each of millions of keys once, and the next morning the cache hit rate has collapsed. Which eviction policy would have resisted that? hard Multiple choice
21. Even with delete-on-write, cache-aside can end up caching a stale value. How, and what bounds the damage? hard Flashcard