On this page
Define the measurement boundary
IP bytes, Ethernet frame sizes and actual link occupancy are different measurements. Record the layer, direction and interval used for every result. At the same bit rate, smaller packets require more processing decisions per second. Results obtained with large packets therefore do not establish small-packet performance.
Change one constraint at a time
On an isolated, authorized test bench, vary frame sizes and then flow counts while keeping the filtering configuration fixed. Record software versions, receive queues, CPU placement and active rules. Include representative legitimate traffic and observe loss, delay and new-session establishment rather than reporting only discarded traffic.
Locate the limiting component
Compare generator, interface and receiver counters. Missing received packets can reflect a generator or intermediate-link limitation rather than the filter itself. Publish results with their test conditions and limitations. The companion article explains the units; this guide concerns reproducible comparison, without presenting laboratory throughput as a guaranteed production capacity.
