Knowing normal before trouble
"The server feels slow" is impossible to act on without a baseline. Record what healthy looks like now, and later you can prove what changed.
Capture normal
Keep a rolling record of the figures that describe health: load and CPU, request latency and throughput, disk IO and queue depth, memory and swap. The point is the shape over a normal week — the daily peak, the quiet hours — not a single snapshot. Monitoring and Prometheus and Grafana are how you keep it.
Baseline, then compare
With a baseline, a regression is obvious: latency up a third since the deploy, IO wait double since the weekend. Without one you are arguing from memory. Re-baseline after a significant change so you are always comparing against current normal, not last year's.
Baseline before you benchmark
A benchmark tells you what the hardware can do; a baseline tells you what your workload actually does. You want both — the gap between them is your headroom, and watching it is the heart of capacity planning.
Still stuck? Mail support@novogara.com — an engineer answers, at any hour. Back to the knowledge base