Uncovering the Light Network Load Performance Penalty of the Network Link Outlier Factor (NLOF)

This thesis evaluates the effectiveness of the Network Link Outlier Factor with Most Likely Link (NLOF: MLL) algorithm under varying network load conditions. Repeated simulation experiments using Mininet were conducted for four different network-wide load levels: 100 Mbps, 500 Mbps, 1 Gbps, and 5 Gbps. Using statistical inference, our experimental results indicate that NLOF: MLL is ineffective under light load conditions (i.e., 100Mbps load) due to the limited network flow data available for its learning process. This limitation highlights a key challenge in applying the algorithm to lightly loaded networks. A preliminary algorithm was proposed to address this light-load performance penalty using synthetic traffic generation. This algorithm demonstrates promising results toward improving the performance of NLOF: MLL. Future work will continue to explore this use of synthetic traffic.