This thesis addresses quality of service measurement in the ARNES access network, where diagnostics are more demanding than in the backbone due to the involvement of third-party providers. The work defines the network hierarchy (core, backbone, access, campus) and fundamental QoS metrics (latency, jitter, packet loss, throughput). A comparison of tools (Ping, MTR, Speedtest-cli, perfSONAR) justifies the choice of iPerf3 due to separate TCP/UDP measurement and support for parallel streams.
The practical part presents an automated system: iPerf3 running in GuestShell on Cisco C1161 routers, orchestrated via Ansible (SCP transfer of the binary, server-side setup, three test modes - normal, reverse, bidirectional). Measurements showed asymmetry between directions, attributed to CPU limitations during TCP transmission, with proposed improvements (parallel streams, kernel parameter tuning).
The result is a foundation for fast, automated quality of service measurement at the edge of the ARNES network without additional server infrastructure at each location.
|