Tinker12:30, mid-sale. Today: 82% hit rate, 162 trips, 324 stale reads.39 from shelves · 6 to the depot · 6 trips · 36 stale
Chapter 7 of 7
Set Larkspur’s cache policy
Everything is on the bench: a TTL per kind, purge and versioned names. The owner has set how stale each kind may be.
This chapter: Which policy keeps the most requests on the shelves within the limits?
About this chapter
You will learn
- How TTLs, purges and versioned names work together
- How to read a day of traffic kind by kind
- Why more than one policy can be equally good
Words you’ll meet
- Cache policy
- The rules for each kind of content: how long to keep it, and how changes reach the shelves.
Step 1 of 2
The owner’s limits: no stale prices or photos, and stock stale on at most 15% of reads.
Set the best .
82% hit rate OVER A LIMIT
| Kind | Hit rate | Stale reads | Limit |
|---|---|---|---|
| Prices | 82% | 33% | 0% |
| Photos | 80% | 2% | 0% |
| Stock counts | 82% | 73% | 15% |
Request coalescing stays on, from chapter 5.
A simulated day: Larkspur’s own traffic and planning numbers, not a measurement of any real network or CDN.