Perancangan dan Evaluasi Konvergensi Failover OSPFv2 Multi-Area untuk Jaringan Multi-Kampus Berbasis SD-WAN

Description

Measurement dataset supporting a study on OSPFv2 multi-area failover convergence for an SD-WAN-based multi-campus network (Universitas Dian Nusantara). Convergence was measured at sub-second resolution (ping -D -i 0.2 -c 300; Tconv from epoch-timestamp gaps) on GNS3 with Cisco IOS 15.2(4)S5 and Alpine Linux (iperf3). Four routing scenarios (floating static, single-area OSPF default, multi-area OSPF default, tuned multi-area OSPF) were tested under two failure modes (link-down and blackhole), 30 repetitions each (n = 30), plus TIPHON-based steady-state QoS. Includes per-run convergence data with timestamps, descriptive statistics, QoS results, an evidence checklist, and Python scripts to reproduce the statistics (Kruskal-Wallis + Dunn post-hoc). Key finding: under blackhole failure, timer tuning reduces convergence from ~40-43 s to ~8.3 s (~80% faster); under link-down all scenarios recover in ~6 s. License: CC BY 4.0.

Steps to reproduce

A. Environment GNS3 with a Cisco IOS 15.2(4)S5 image (C7206VXR/Dynamips) on three routers (TJD, CBB, GV) following the topology; two Alpine Linux container hosts (with iperf3 and ping/iputils installed). Apply each router’s running-config for the scenario under test (A: floating static + IP SLA/track; B: single-area OSPF, default timers 10/40; C: multi-area OSPF, default timers; D: multi-area OSPF, tuned timers Hello 1 s/Dead 3 s). Backup path: IPsec IKEv2 tunnel (AES-256-CBC, SHA-256, DH group 14). Verify the initial state: show ip ospf neighbor (FULL), show ip sla statistics 100 (OK), show track 10 (Up), show crypto ikev2 sa (READY). B. Convergence measurement (repeat 30× per scenario × mode) 4. From the source host run: ping -D -i 0.2 -c 300 10.3.1.2 | tee konv_<Scenario>_<Mode>_run<NN>.txt 5. While the ping is running, inject a primary-path failure: Mode 1 (link-down): shutdown the primary-path interface (physical-layer detection). Mode 2 (blackhole): enable a GNS3 Packet filter → Drop 100% on the primary link (interface stays up; detection only via the timer). Wait for traffic to recover over the backup path (IPsec Tunnel0), then restore the primary path before the next run. C. Compute Tconv 7. Run python scripts/hitung_tkonv_dari_ping.py in the folder containing the konv_*.txt files. The script finds the longest time gap and computes Tconv = timestamp of the first successful packet after the failure − the last successful packet before the failure. Copy the results into data/konvergensi_wide.csv (or use the provided file). D. Steady-state QoS 8. Measure under steady state (no failure) for both the primary path and the IPsec backup path at offered loads of 50/100/200 Kbps: use iperf3 for throughput/jitter/loss and ping -c 100 for delay; classify with the TIPHON standard (see data/qos_steadystate.csv). E. Statistical analysis (reproduce the paper’s figures) 9. pip install pandas scipy scikit-posthocs 10. cd scripts && python analisis_kruskal_wallis.py 11. The output reproduces the descriptive statistics (mean/median/SD) and the Kruskal-Wallis test with Dunn post-hoc per failure mode, matching Table 8 and the discussion in the manuscript.

Institutions

• Universitas Dian Nusantara

  Jakarta, Jakarta

Categories

Licence