Compute-resource capacity is determined by the workload and its performance targets, not simply by dividing host RAM by a plan size. Monitor sustained and peak behavior before allocating more SolusVM servers. A lightly used test guest does not represent a busy database or build server.
IN THIS ARTICLE
Gather the details you will need
Record physical CPU, memory, storage layout, network capacity and existing allocations. Define acceptable latency and reserve capacity for the host, backups, migrations and any HA commitments. Keep utilization history long enough to include scheduled jobs and peak demand.
01 Diagnose compute-resource capacity in SolusVM 2
- Compare provisioned resources with actual host consumption. Track CPU saturation and scheduling delay, available memory and swapping, storage latency/capacity, and network throughput or loss.
- Correlate spikes with backup windows, guest creation, antivirus scans or customer workloads. Inspect host OOM events and storage errors rather than relying only on an averaged dashboard graph.
- Check thin-pool data and metadata separately where applicable. Account for snapshots and temporary migration files that can consume space beyond the live guest allocation.
- Test representative workloads in an isolated capacity exercise. Increase load gradually while measuring existing guest service levels; do not deliberately stress paying production workloads to failure.
- Set alert thresholds and provisioning limits with operational headroom. Stop placement on constrained resources and plan migration or added capacity before users encounter outages.
How many VPSs can this host run?
There is no universal count. Plan size, actual demand, storage latency, isolation requirements and reserved recovery capacity all matter. Establish a tested limit and revisit it as workloads change.
Confirm the expected result
Confirm alerts are delivered and contain the correct host identity. Review performance during normal peaks and a backup cycle. Check that the remaining resources could meet the intended maintenance or failover plan, not only today's steady-state usage.
If the problem continues
If high load has no corresponding CPU use, investigate disk waits and blocked processes. If memory appears free but guests pause, inspect host pressure, swapping and storage latency. Avoid applying a vendor example oversubscription ratio as a universal safe limit; reduce allocations or move workload according to observed constraints.