In the performance evaluation of communication networks, packet-level simulation entails enormous computational costs under conditions such as large-scale networks or numerous communicating flows, leading to issues where execution time exceeds practical limits. In contrast, flow-level simulation, which approximates traffic as a continuous fluid, can mitigate the increase in computational complexity associated with the growing number of packets. We have previously proposed a method to accelerate simulations by circulating a token, which carries information regarding the network state, along the path traversed by a flow during the simulation. However, existing token-based simulators assume a fixed step size and perform state updates at constant intervals, even during periods where state changes are minimal. Consequently, redundant computations occur relative to the required accuracy. In this study, we propose a method that introduces dynamic step size control to a token-based flow-level simulator. To address the structural issue of token history omissions caused by step size alterations, we propose a hybrid step-doubling method that estimates missing token information through token interpolation and direct calculation of internal states, thereby autonomously controlling the step size. Evaluation experiments conducted in a simple environment using a dumbbell network topology with a single bottleneck confirmed the effectiveness of the proposed method. Compared to the conventional fixed step size method, the simulation execution time was reduced by approximately 90% to 97%, and memory consumption was reduced by approximately 25% to 70%.
Base-T1S Ethernet is a promising technology for integrating low-rate electronic control units (ECUs) in invehicle and industrial networks, because it supports multidrop communication over a single-pair physical layer while preserving an Ethernet-based architecture. Its Physical Layer Collision Avoidance (PLCA) mechanism assigns deterministic transmission opportunities to nodes in a round-robin manner. IEEE 802.1Qav Credit-Based Shaper (CBS) is widely adopted in TSN/AVB networks to regulate class-based bandwidth and suppress traffic burstiness, and can be deployed locally within each ECU before frames are forwarded to the PLCA MAC-PHY. Existing CBS analyses typically assume full-duplex point-to-point Ethernet links, where an eligible queue, once selected by the local scheduler, can receive continuous transmission service at the link rate. This assumption does not directly hold in PLCA-based multidrop networks, where an ECU can transmit only during its assigned transmission opportunities. For control, diagnostic, and safety-related traffic, average delay or finite simulation traces are insufficient to certify whether timing requirements are always met; designers instead need deterministic upper bounds on the worst-case delay. However, in PLCA-based 10Base-T1S networks, the service available to CBS-shaped traffic is determined by the interaction among CBS credit evolution, PLCA access timing, and local non-preemptive priority scheduling. Therefore, these mechanisms must be modeled jointly to derive such bounds. This paper develops a network-calculus model for CBS-shaped traffic over PLCA-based 10Base-T1S networks. Numerical examples and simulation results demonstrate how CBS parameters and PLCA configurations influence the resulting delay bounds.
Taiki Nonaka, Han Nay Aung, Yasuhiro Yamasaki et al.· Annual International Compute...· 0 citations
Ensuring the reliability of quantum networks requires the accurate evaluation of fidelity-a metric representing link quality-and the selection of high-fidelity links. However, since fidelity estimation consumes a large number of measurements, an efficient method for identifying high-quality communication links with limited resources is desirable. Existing studies have primarily focused on identifying high-fidelity links between a pair of nodes, leaving adaptive resource allocation insufficiently explored in multi-destination environments where destinations differ in network importance (e.g., communication demand or the quantum memory capacity of nodes). In this study, we propose DaTopLinks (Demand-aware Top- $K$ HighFidelity Links), a method for efficiently identifying the top- $K$ destinations and their best link for each selected destination according to a utility function that combines destination importance and link fidelity. The algorithm introduces a dual-criterion link elimination mechanism that simultaneously performs intradestination link elimination and top- $K$ destination elimination. It also incorporates an early confirmation mechanism that allows early termination of measurements for destinations once their inclusion in the top- $K$ set and the identification of their best link are statistically guaranteed. In our theoretical analysis, we derive an upper bound on the sample complexity based on an effective gap that captures both inter-destination utility differences and intra-destination fidelity differences. Simulation results demonstrate that the proposed method functions effectively under depolarizing, dephasing, and bit-flip noise models.
Shun Yamachika, Yuto Kakihara, Shota Inoue et al.· Annual International Compute...· 0 citations
Ensuring deterministic and reliable communication is essential for in-vehicle networks supporting autonomous driving and safety-critical functions. Time-Sensitive Networking has emerged as a key enabler for such systems. Among its mechanisms, the IEEE 802.1Qcr Asynchronous Traffic Shaper (ATS) offers fine-grained traffic control without requiring global time synchronization. However, the practical deployment of ATS in Automotive Ethernet networks remains challenging due to the difficulty of parameter configuration. The performance of ATS strongly depends on the appropriate setting of key parameters such as the Committed Information Rate (CIR) and Committed Burst Size (CBSz), which are highly sensitive to both network topology and traffic workload. Conventional approaches relying on static configuration or empirical tuning may face difficulties in ensuring QoS when network conditions change. This paper proposes a method for automated, high-precision optimization of ATS parameters in automotive networks. We analyze the impact of key parameters—CIR and CBSz—on delay and frame loss, and develop a machine learning model to select optimal settings under dynamic traffic conditions. Our results reveal that proper ATS parameter configuration is essential for deterministic latency and reliability in Automotive Ethernet networks. CIR mainly governs bandwidth, affecting queuing delay and frame loss, while CBSz balances delay reduction against burst-induced congestion. Furthermore, tree-based ensemble models such as LightGBM and Gradient Boosting achieved high prediction accuracy and QoS satisfaction under varying traffic conditions.
Taisei Isobe, Han Nay Aung, Yasuhiro Yamasaki et al.· Annual International Compute...· 0 citations