How do I measure network traffic without running a disruptive speed test?
Read existing traffic counters or short interval statistics before launching a bandwidth-saturating test. Interface traffic reflects the selected layer and direction; bonds, bridges, guests and provider billing systems can account differently. A local NIC rate is not a guaranteed measure of useful application throughput or billable transfer.
Before you start
- sysstat is installed for sar and interface monitoring is authorised.
Instructions
1. Collect a bounded interface sample
Where sysstat is installed, sample network devices during the actual workload. Identify the physical uplink and any higher-level interfaces to avoid counting the same traffic twice.
sar -n DEV 1 5This reports five one-second intervals. Check the displayed units and interface names before calculating bandwidth.
2. Compare cumulative counters if needed
Record the timestamp and counters twice over a known interval. For byte counters, average bits per second is the byte difference multiplied by eight and divided by elapsed seconds.
date -u -Is
ip -s link show dev enp1s0Account for counter resets and the exact interface layer. Compare sustained rates with the service allowance and provider monitoring.
Check the result
- The measurement window and units are explicit, traffic is attributed to the correct interface and application performance is assessed separately.
Troubleshooting
- The provider dashboard shows a different value
- Check sampling intervals, direction, encapsulation and billing definitions before assuming either measure is wrong.
- A full-speed benchmark is required
- Coordinate the destination, duration, traffic allowance and production impact before generating extra load.
Official references
Use the documentation for your installed version when a flag, service name or platform feature differs.