COMPLEMENTARY, NOT SUBSTITUTIVE: LOAD-DEPENDENT INTERACTION EFFECTS OF FIBRE-OPTIC INFRASTRUCTURE AND SOFTWARE-DEFINED NETWORKING ON ISP NETWORK PERFORMANCE IN A KENYAN RESEARCH-AND-EDUCATION NETWORK

Abstract

Fibre-optic infrastructure and Software-Defined Networking (SDN) are each independently documented to improve network performance, yet whether their combined effect is simply additive or functionally interactive has not been empirically tested in an African ISP context. This paper reports that test, using KENET, Kenya's national research and education network, as the empirical case. Controlled network emulation (Mininet 2.3, NS-3.38) on a twelve-node, three-tier replica of KENET's backbone compared a static-routing baseline against an SDN-integrated condition across a graduated traffic-load range (40–95 percent), cross-validated against KENET's own monitoring data (mean divergence 9.8 percent) and corroborated through 23 practitioner interviews (76.7 percent response rate). SDN deployment scope, entered as a continuous predictor, explained 76 percent of the variance in latency (R² = 0.76, F(1, N) = 48.6, p < .001). Critically, the combined benefit was not additive: latency improvement rose from 54.8 percent at 40 percent load to 65.4 percent at 95 percent load, P95 throughput improvement rose from 18.3 to 51.1 percent over the same range, and modelled annual uptime rose from 99.72 to 99.95 percent under automated failover. This load-dependent pattern was robust under sensitivity analysis (48.9–66.2 percent) and independently corroborated by practitioner-reported estimates (50–60 percent). H₀₃ was rejected in favour of H₁₃: fibre-optic infrastructure and SDN function as complementary rather than substitute investments, with SDN's marginal value concentrated precisely where fibre capacity is under the greatest strain. The findings imply that ISPs should sequence SDN rollout to the most congested segments first, capturing disproportionately greater marginal benefit than uniform deployment.