Chapter 6: An ACO Algorithm for Data Network Routing

6.3 The Experimental Settings

• NSFnet: 14 nodes, 21 bidirectional links, 1.5Mbit/s, 4 - 20ms propagation delays, well-balanced
• NTTnet: 57 nodes, 162 bidirectional links, 6Mbit/s, 1 - 5ms propagation delays, not well-balanced
• A session is characterized by the number of transmitted packets, their size, and the interarrival distribution
• Temporal Distribution: Poisson (P), Fixed (F), Temporary (TMPHS)
• Spatial Distribution: Uniform (U), Random (R), Hot Spots (HS)
• Performance Evaluation: Throughput, Delay Distribution
• Competing algorithms: OSPF (static, link-state), SPF (adaptive, link-state), BF (adaptive, distance-vector), Q-R (adaptive, distance-vector), PQ-R (adaptive, distance-vector), Daemon (adaptive, optimal routing)
• Box 6.1: experimental settings for AntNet

6.4 Results

• 10 trials, 1000 virtual seconds per trial, 500 preliminary simulation seconds
• MSIA: mean session interarrival time distribution of Poisson case
• MPIA: mean of the packet interarrival time distribution
• HS: number of hot spots
• MPIA-HS
• AntNet data packets are routed probabilistically
• Other algorithm data packets are routed deterministically
• NSFNet: Figures 6.6, 6.7, 6.8, 6.9
• NTTNet: Figures 6.10, 6.11, 6.12, 6.13
• Routing Overhead: Table 6.2, Figure 6.14

6.5 AntNet and Stigmergy

• r is one of the factors that influences pheromone modification. It can be a constant or a function of the ant's trip time and local statistical models. Figure 6.15