Table 2.
Performance metrics for evaluating DQL in SDN.
| Metric | Description | Significance | Example findings/references |
|---|---|---|---|
| Throughput | Total data successfully delivered over time | Reflects link utilization and efficiency | DQL improved throughput by 18–22% vs. ECMP (Fu et al., 2020) |
| Packet Loss Rate | % of packets dropped due to congestion or link failure | Indicates network reliability | Loss reduced by ~15% under mixed workloads (Zhang et al., 2023) |
| Avg. End-to-End Delay | Time for a packet to travel from source to destination | Critical for real-time and QoS-sensitive apps | Temporal DQL cut delay by ~12% (Sharma et al., 2024) |
| Convergence Time | Time required for DQL to learn a stable routing policy | Demonstrates learning efficiency | DRL-R converged 30% faster than Q-learning (Liu et al., 2021) |