Tiering hot and cold data
Most data is cold most of the time. Paying flash prices to store all of it is waste; the trick is keeping only the hot working set on the fast tier.
Split by how often it is touched
A small fraction of your data serves most of the requests — the active database, recent files, the current index. That belongs on NVMe. The long tail of archives, old logs and backups belongs on large, cheap drives where latency does not matter.
Manual tiers beat clever ones
The simplest tiering is explicit: put the hot paths on a fast volume and the bulk on a slow one, and let the application decide what goes where. It is predictable and easy to reason about. Automatic tiering (LVM cache, ZFS special and cache devices) helps when the hot set shifts on its own, at the cost of some unpredictability.
Build the box for it
A tiered machine mixes a little flash with a lot of spinning capacity — exactly what a storage-heavy server is for. Tell us the size of your working set versus your archive and we spec the two tiers to match.
Still stuck? Mail support@novogara.com — an engineer answers, at any hour. Back to the knowledge base