cloudpatterns.appClassroomKatasWhiteboardBlog
Cache LineLesson

Cache basics — hit ratio, TTL, eviction

A cache's hit ratio isn't a setting — an entry earns hits only if it's read again while the cache still holds it. What fits caps the ratio, a TTL trades freshness for hits, and when the cache is full an eviction policy guesses what to drop.

12 minEasyStation: Cache basics: hit ratio, TTL, eviction

Members lesson

Membership isn’t open yet. The free lessons are open — 16 of them.

Start a free lessonSee the map
  1. Memory vs disk vs network
  2. Cache basics: hit ratio, TTL, eviction
  3. Cache-aside
  4. Read-through & write-through
Cache Line — where this lesson sits

What you'll be able to do

  1. 1Explain what sets a cache's hit ratio — whether the same data is asked for again while the cache still holds it — and work out the load it leaves on the store behind.
  2. 2Explain how a TTL trades freshness against hit ratio, and what happens when an entry expires.
  3. 3Size a cache against the hot set its reads fall on, and choose an eviction policy — LRU, LFU or random — for the access pattern.

Inside the lesson

steps
10
questions
8
simulations
3

You guess first, then run the system and see if you were right.

Step 1 · The setup

At the last stop, Pixel's team saw that memory answers in nanoseconds while the database's disk takes milliseconds — so a copy of busy data in memory, a cache, pays.

Pixel's feed API answers 200 feed reads a second, each a query on the photo database. Readers ask for 1,000 different feeds, but not evenly: the busiest 200 draw 80% of the reads.

The team puts a cache in front of the database, with room for 200 feeds. Once warm, it holds the feeds read most.

Steps 2–10 are for members

More on the Cache Line

  1. Memory vs disk vs network — the latency ladderOpen11 min
  2. Cache-asideFree with an account10 min
  3. Read-through & write-throughMembers11 min
  4. Cache invalidationMembers12 min
  5. Cache stampedeMembers11 min