Name: System Design Masterclass
Availability: InStock

Write-through sends every write to the cache and to the store together, so the cached copy is never behind. That is a real property and it is why the pattern exists.

A sequence diagram between the application, the cache and the database. The application writes the new value to the cache, which writes it to the database too, and only then is the write complete, so the cached copy is never behind. A note reads that the write path now includes the cache. — SequenceFigure 1. The write is not acknowledged until both the cache and the store have it
There is no invalidation to get right, because nothing is ever left stale. And every write now occupies a slot that some other entry was using.

The consistency argument is sound, and this lesson is about the second sentence. A cache is a fixed number of slots, and write-through decides how some of them are spent.

A hand-drawn figure. The top row shows reads over 50,000 keys as bars in a steeply falling Zipf shape. The bottom row shows writes over the same 50,000 keys as ten bars of equal height. A note reads that the cache holds 5,000 slots and both of these are competing for them. — SketchFigure 2. Reads are concentrated and writes are spread out, and they share one cache
A user editing their own profile, an order being updated, an inventory count changing. Every one of those is roughly as likely as any other, and almost none of them is a popular item.

Where this shows up in interviews

Where this shows up in interviews

Write-Through Cache

The rest of this lesson

Continue this lesson

Every Write Puts Something In The Cache

What Happened To The Read Hit Rate