研究目的
To compare the performance of centralized and distributed spectrum management techniques in elastic optical networks, focusing on connection success rate, connection establishment time, and capacity requirement under different traffic conditions.
研究成果
Centralized network operation mode (C-NOM) outperforms distributed mode (D-NOM) in busy hour traffic with 100% connection success rate and efficient spectrum utilization, but it requires prior knowledge of network state. D-NOM is more scalable and performs better under normal traffic conditions with less resource contention. Backward Reservation Protocol (BRP) provides better performance than Forward Reservation Protocol (FRP) in D-NOM. The findings guide network architects in choosing appropriate control-plane strategies based on traffic conditions.
研究不足
The simulations did not explicitly consider guard bands, which could affect spectrum efficiency. The propagation delay was assumed constant for simplicity, and the study focused on generic algorithms without advanced optimization techniques. The centralized mode relies on predefined optimal data, which may not be practical in dynamic environments.
1：Experimental Design and Method Selection:
A network simulator was developed in C++ to model EONs with GMPLS control plane, using Forward Reservation Protocol (FRP) and Backward Reservation Protocol (BRP) signaling methods, along with First Fit (FF) and Random Fit (RF) RSA algorithms. The design includes simulations for both centralized (C-NOM) and distributed (D-NOM) network operation modes under busy hour (parallel arrivals) and normal (Poisson arrivals) traffic conditions.
2：Sample Selection and Data Sources:
Two network topologies were used: Finland topology with 12 nodes and 19 fiber links, and Germany topology with 17 nodes and 26 fiber links. Each link has 160 frequency slots of 25 GHz, totaling 4 THz spectrum capacity. Data from realistic case studies [33] were referenced for centralized mode parameters.
3：List of Experimental Equipment and Materials:
The simulation tool was developed in C++ with a GUI for input. No physical equipment was used; it is a software-based simulation.
4：Experimental Procedures and Operational Workflow:
For busy hour simulations, a high arrival rate of 1000 connections per second was used with exponentially distributed service times. For normal traffic, Poisson arrivals were simulated. The RSA algorithms (FF and RF) were applied with FRP and BRP signaling to compute paths and allocate spectrum. Each simulation run was averaged over 10 consecutive runs.
5：Data Analysis Methods:
Performance metrics such as connection success rate, average connection establishment time, and maximum frequency slot value were analyzed. Statistical analysis included linear fitting for blocking probability trends.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
