Write-through
Cache policy: every write goes to both cache and underlying media at the same time. Read-cache benefit retained; write performance is the same as if there were no cache. Safe against power loss; slow.
Use when: RAID controller without BBU/supercapacitor, hardware RAID with battery fail, critical metadata where data loss is unacceptable (filesystem journals). Default fallback when safety > performance. Filesystem cache: often write-back in Linux page cache but fsync forces effective write-through to the underlying device. RAID5/6: write-through avoids the write-hole problem (parity corruption on power loss mid-stripe-update). Modern enterprise SSDs with PLP — write-back is safe. Consumer-without-PLP — write-through mode is better for critical workloads.