International Journal of Network Management最新文献

The core of IPv6 active address discovery lies in identifying operational network devices, which serve as the foundation for various network applications. However, the vast address space of IPv6 presents significant challenges to network discovery, including inefficiencies of traditional scanning tools (such as ZMap and Masscan), uncertainties arising from dynamic address allocation, sparsity in address distribution, and limitations of current discovery techniques in regions lacking seed addresses. To address these challenges, researchers have developed various address discovery methods aimed at maximizing the identification of active IPv6 addresses within constrained resource budgets. This paper summarizes and analyzes existing active address discovery methods and evaluates their performance in real-world environments. Our work includes classifying detection techniques and outlining the current research landscape for address discovery in regions without seed addresses. Through experiments conducted in real network environments, we evaluate and contrast different probing algorithms using four performance metrics: hit rate, hierarchical prefix coverage, aliased addresses, and address discovery rate. Furthermore, we discuss the challenges faced by current IPv6 active address discovery methods in practical applications. Despite numerous advancements, the field of IPv6 active address discovery still requires further research and exploration.

This paper investigates the application of the Advantage Actor–Critic (A2C) reinforcement learning model in cryptocurrency trading and portfolio management, focusing on the impact of price prediction accuracy and data granularity on model performance. The highly volatile and 24/7 operating cryptocurrency market presents unique challenges that require sophisticated trading strategies. Our study examines how varying levels of noise in price predictions affect the A2C model's ability to manage a diversified portfolio and generate returns, revealing that higher prediction accuracy leads to improved performance. Additionally, we explore the role of data granularity, finding that overly fine-grained data introduce excessive noise that impairs the model's performance, whereas data processed at 6- and 12-h intervals optimize trading efficiency and profitability. These findings show the importance of maintaining sufficient prediction accuracy and selecting appropriate data granularity to enhance the effectiveness of reinforcement learning models in cryptocurrency trading, providing valuable insights for developing robust artificial intelligence (AI)-driven trading strategies in volatile markets.

Bitcoin is the first decentralized digital currency built upon blockchain technology and has facilitated financial innovation and technological advancement. However, the growth in its popularity has uncovered the limitations of the blockchain technology, leading to issues such as transaction congestion, high transaction fees, and delayed confirmations. This has motivated the exploration of the Lightning Network, a payment channel network designed to enhance Bitcoin's scalability and transaction efficiency. Serving as a layer-2 protocol, the Lightning Network supports off-chain transactions, significantly reducing fees and enabling swift and cost-effective micro payments. This paper comprehensively examines the technology, open source implementations, and relevant research of the Bitcoin Lightning Network, covering key categories such as “network analysis,” “network security,” “payment routing,” “rebalancing,” and “network design and applications.” With the Lightning Network's increasing importance in the blockchain and financial domains, this survey provides a valuable resource for researchers seeking to explore this technology in greater depth.

In cloud computing environments, user requests often lead to varying load conditions across the system, resulting in underloaded, overloaded, or balanced states. Both underloading and overloading can cause system inefficiencies, including increased power consumption, prolonged execution times, and higher machine failure rates. Therefore, effective load balancing (LB) becomes a critical aspect of task scheduling in cloud systems, whether at the level of virtual machines (VMs) or independently. To address these challenges, this paper proposes the Chronological Kookaburra Optimization Algorithm (ChKOA) for efficient LB in cloud computing (CC). The proposed ChKOA is the combination of chronological concept with the Kookaburra Optimization Algorithm (KOA). Initially, tasks are assigned to VMs in a round-robin manner. Based on specific VM parameters, the VMs are classified into overloaded and underloaded categories using deep embedded clustering (DEC). Tasks in overloaded VMs are prioritized and redistributed to underloaded VMs, considering factors such as supply, demand, capacity, predicted load, and key Quality of Service (QoS) metrics, including resource availability and reliability. Load prediction is performed using a Deep Residual Network (DRN). Simulation results demonstrate that the proposed ChKOA achieves a balanced load of 0.535, capacity utilization of 0.954, resource availability of 0.954, reliability of 0.936, and a computational cost of 0.327 s.

The Quick UDP Internet Connections (QUIC) protocol provides a secure, reliable, and low-latency communication foundation for HTTP/3. Connection migration is a key technology of QUIC. When the IP/Port of a connection changes, the connection ID is used to maintain a secure and uninterrupted connection. However, current connection migration is passive, designed to support mobile handover and weak network environments. In this paper, we propose proactive connection migration for QUIC (PCM-QUIC), which combines connection migration and online path selection, enabling QUIC to select the best quality transmission path while maintaining the connection. First, PCM-QUIC integrates the exploration of network quality across different paths into multiple user request actions. Then, considering response completion time and jitter, PCM-QUIC identifies the optimal access path for the current Internet service through online learning. In addition, we propose an upper confidence bound-based path selection algorithm with the goal of minimizing the confidence upper limit of the path reward. Experimental results show that, compared with the original QUIC, PCM-QUIC reduces the average response completion time by up to 59.43%.

In contemporary networking infrastructure, the integration of software-defined networking (SDN) concepts with traditional networking methodologies has given rise to hybrid software-defined networking (HSDN); as the current network infrastructure evolves, it combines the programmability of SDN with the established protocols of traditional network infrastructure. This paradigm shift introduces novel opportunities and challenges in traffic engineering (TE), necessitating innovative solutions to enhance network performance, resource utilization, and efficiency. In this study, we examined the challenges associated with TE in a hybrid SDN environment, where traditional network devices coexist alongside SDN nodes; we investigated the routing optimization of traffic engineering in a migrated hybrid network and formulated the problem as a mixed integer non-linear programming model. We proposed a heuristic algorithm (H-STE) that optimizes both the OSPF weight setting and the splitting ratio of SDN nodes in the hybrid environment. Extensive evaluations were conducted using real network topologies to validate our method. The results demonstrate that a 30% deployment ratio of SDN nodes significantly improves traffic engineering performance. Specifically, the Maximum Link Utilization (MLU) stabilizes at this ratio, indicating near-optimal network efficiency. This research provides valuable insights for researchers, practitioners, and network architects navigating SDN, hybrid SDN, and traffic engineering optimization.