1 / 5 Cache Patterns & Write Policies

Import Storm

A nightly job bulk-inserts about 40 million audit records. Fewer than 2% of them are ever read back, but since the job started running, the cache hit rate on the user-profile keys has collapsed. Which write policy fixes this?

A

Refresh-ahead: reload each audit key asynchronously shortly before its TTL expires, so no user request ever pays for reloading an audit record

B

Write-around: send the audit inserts straight to the database and let the cache fill only for records that are actually read

C

Write-through: keep writing the audit records into the cache as they are inserted, so any later read of an audit row is served without a query

D

Write-behind: buffer the audit inserts in the cache and flush them to the database in batches, so the import finishes with fewer round trips