Add explicit recommendation to use NVMe / SSD for DB storage (#188)

Signed-off-by: nscuro <nscuro@protonmail.com>

Author: nscuro

docs/guides/administration/configuring-database.md

@@ -52,6 +52,11 @@ without being impacted by other applications.
 
 For smaller and non-critical deployments, it is totally fine to run everything on a single machine.
 
+**You should use fast disks for storage**, ideally NVME, but at the very least SSDs. Unless your entire working
+set fits into memory, PostgreSQL needs to perform random disk access, which is slow on HDDs due to seek latency.
+Local disks give you the most predictable performance. Network-attached storage also works well, provided you
+provision guaranteed IOPS and avoid tiers that throttle throughput under sustained load.
+
 #### Upgrades
 
 Follow the [official upgrading guide]. Be sure to select the version of the documentation that corresponds to the

docs/guides/administration/deploying-to-production.md

@@ -50,6 +50,10 @@ Run the database on a dedicated host. Co-locating PostgreSQL with API server ins
 compete for the same CPU, memory, and I/O, and that contention surfaces as unpredictable latency under
 load.
 
+Use fast disks, ideally NVME but at least SSDs. Once the working set outgrows memory,
+PostgreSQL falls back to random disk access, which HDDs handle poorly.
+See [Configuring the database](configuring-database.md) for storage guidance.
+
 When sizing the database host, account for the per-instance connection pool. PostgreSQL's `max_connections`
 must cover the sum of all pools across the cluster, plus headroom for migrations, backups, and
 administrator sessions: