DRR-MDPF: A Queue Management Strategy Based on Dynamic Resource Allocation and Markov Decision Process in Named Data Networking (NDN)

Abstract

Named Data Networking (NDN) represents a transformative shift in network architecture, prioritizing content names over host addresses to enhance data dissemination. Efficient queue and resource management are critical to NDN performance, especially under dynamic and high-traffic conditions. This paper introduces DRR-MDPF, a novel hybrid strategy that integrates the Markov Decision Process Forwarding (MDPF) model with the Deficit Round Robin (DRR) algorithm. MDPF enables routers to intelligently predict optimal forwarding decisions based on key metrics such as bandwidth, delay, and the number of unsatisfied Interests, while DRR ensures fair and adaptive bandwidth allocation among competing data flows. The proposed method models each router as a learning agent capable of adjusting its strategies through continuous feedback and probabilistic updates. Simulation results using ndnSIM demonstrate that DRR-MDPF significantly outperforms state-of-the-art strategies including SAF, RFA, SMDPF, and LA-MDPF across various metrics such as throughput, Interest Satisfaction Rate (ISR), packet drop rate, content retrieval time, and load balancing. Notably, DRR-MDPF maintains robustness under limited cache sizes and heavy traffic, offering enhanced adaptability and lower computational complexity due to its single-path routing design. Furthermore, its multi-metric decision-making capability enables more accurate interface selection, leading to optimized network performance. Overall, DRR-MDPF serves as an intelligent, adaptive, and scalable queue management solution for NDN, effectively addressing core challenges such as resource allocation, congestion control, and route optimization in dynamic networking environments.