Reducing load.
WHAT CACHING DOES
Stores the result of expensive work so it need not be repeated.
WHAT TO CACHE
Query results that are expensive and rarely change Rendered fragments Computed aggregates Reference data
WHAT NOT TO CACHE
Anything that must be current Data specific to one user, unless keyed properly
WHY THAT SECOND POINT MATTERS
A cache keyed carelessly serves one user's data to another.
WHAT THAT IS
A serious breach, and it has happened to many products.
WHAT TO ALWAYS INCLUDE IN A CACHE KEY
Everything the result depends on, including the user or tenant.
WHAT INVALIDATION MEANS
Removing cached data when the underlying data changes.
WHY IT IS THE HARD PART
Knowing everything a change affects is genuinely difficult.
WHAT THE SIMPLEST APPROACH IS
Short expiry times.
WHY IT WORKS
Staleness is bounded, and no invalidation logic is needed.
WHAT TO MEASURE
Hit rate.
WHAT A LOW RATE MEANS
The cache is not helping and may be costing.
WHAT TO BE CAREFUL WITH
A cache that becomes essential rather than an optimisation.
WHY
Its failure then takes the application down.
WHAT TO TEST
Behaviour with the cache empty or unavailable.
