COST239

该数据集采用了COST239网络拓扑与NSFNet结合的方式，以评估不同强化学习代理在光网络资源分配方面的表现。在数据集中，假设流量特性是完全动态的，连接请求以泊松过程到达，而连接持续时间则遵循负指数分布。该数据集的规模是动态的，包含100个FSUs和每条链路3个核心，任务涉及路由、调制、光谱以及核心分配（Rmsca）。

This dataset combines the COST239 network topology and NSFNet to evaluate the performance of various reinforcement learning agents in optical network resource allocation. In this dataset, traffic characteristics are assumed to be fully dynamic: connection requests arrive following a Poisson process, and connection durations follow a negative exponential distribution. This dataset features a dynamic configuration, including 100 FSUs and 3 cores per link, with the task covering routing, modulation, spectrum, and core assignment (Rmsca).