Prometheus sizing is driven by the volume and cardinality of time-series data rather than by the number of dashboards. Target count, scrape interval, label design, retention, and recording rules all affect RAM and disk.
A small monitoring VPS can be inexpensive, but poor label cardinality can overwhelm much larger servers.
Key takeaways
- 2 GB works for small Prometheus installations; 4 GB provides safer headroom.
- Retention and metric cardinality drive storage and memory.
- Fast SSD or NVMe helps compaction and query responsiveness.
- Long-term retention may belong in remote storage rather than one VPS.
- Monitor Prometheus itself to understand growth.
How much RAM for Prometheus?
A small home lab or handful of servers can fit in 2 GB. Four gigabytes is more comfortable as target count, alert rules, and metric volume grow.
The biggest danger is uncontrolled cardinality from labels that create enormous numbers of unique series.
Storage and retention
Prometheus stores time-series data locally, so retention length directly affects disk usage. More targets and shorter scrape intervals produce more samples.
Set a retention policy deliberately instead of allowing the disk to become the limit.
Query and compaction load
Large range queries, recording rules, and compaction can create CPU and disk spikes. Fast storage helps, but good metric design is more important.
Dedicated CPU is rarely required for small setups but can help larger observability workloads.
When to use remote storage
If months or years of metrics are required, remote-write systems or dedicated long-term storage can scale better than one VPS.
Keep the primary Prometheus instance focused on the retention period needed for operational monitoring.
Frequently asked questions
Is 2 GB enough for Prometheus?
Yes for small installations, though 4 GB gives more headroom for growth.
Why does Prometheus use so much RAM?
Active time series and label cardinality consume memory. High-cardinality metrics can grow resource use rapidly.
Does Prometheus need NVMe?
Fast storage helps writes, compaction, and queries, especially at higher ingestion rates.
Can Grafana and Prometheus share a VPS?
Yes for small setups. Size the server primarily around Prometheus data volume.