How to choose: answering the main questions
How many resources does monitoring consume?
Zabbix for 200 hosts with one-minute intervals requires a modest 4–8 cores and 16–32 GB of RAM, but the database grows: plan for at least 1 TB of SSD storage. Prometheus+Loki on the same infrastructure will require more disk space for logs — estimate 1–2 GB per host per day.
Is an SSD mandatory?
Yes: time-series databases write constantly in small blocks. Leave HDDs for long-term log archives; keep the hot data (7–14 days) on SSD/NVMe.
What about the fault tolerance of the monitoring system itself?
Minimum: daily backup of configuration and database to another machine. For critical environments, a second node with replication is required: monitoring should not be a single point of failure.