📚 Network Mathematical Models and Applications | 网络数学模型及其应用
Mathematical models provide a rigorous framework for understanding, designing, and optimising computer networks. By representing network components and behaviours with equations and abstractions, engineers can predict performance, identify bottlenecks, and ensure reliability before physical deployment. This article explores the core mathematical models used in computer networking and their practical applications.
数学模型为理解、设计和优化计算机网络提供了严谨的框架。通过用方程和抽象表示网络组件与行为，工程师可以在物理部署之前预测性能、识别瓶颈并确保可靠性。本文探讨了计算机网络中使用的核心数学模型及其实际应用。
1. Graph Theory Foundations | 图论基础
A computer network is naturally represented as a graph G = (V, E), where V is the set of nodes (routers, switches, hosts) and E is the set of links. Graph theory supplies tools for analysing connectivity, shortest paths, and robustness. For example, the adjacency matrix A and the incidence matrix allow compact storage and algorithmic manipulation of network structure.
计算机网络天然地表示为图 G = (V, E)，其中 V 是节点集（路由器、交换机、主机），E 是链路集。图论为分析连通性、最短路径和鲁棒性提供了工具。例如，邻接矩阵 A 和关联矩阵允许对网络结构进行紧凑存储和算法操作。
Key graph metrics include degree centrality, betweenness centrality, and clustering coefficient. These metrics identify critical nodes whose failure would fragment the network, guiding the placement of redundant links and monitoring systems.
关键图指标包括度中心性、介数中心性和聚类系数。这些指标能够识别出那些故障会导致网络分裂的关键节点，从而指导冗余链路和监控系统的部署。
2. Network Topology Models | 网络拓扑模型
Topology defines the physical or logical arrangement of nodes and links. Common models include bus, star, ring, mesh, and tree. Mathematically, each topology imposes different constraints on path lengths and fault tolerance. For instance, a star network has a central hub: the failure of the hub disconnects all nodes, whereas a full mesh survives any single-link failure.
拓扑定义了节点和链路的物理或逻辑排列方式。常见模型包括总线型、星型、环型、网状和树状。在数学上，每种拓扑对路径长度和容错性施加了不同约束。例如，星型网络具有中心集线器：集线器故障会使所有节点断开，而全网状网络能够承受任意单条链路故障。
Topological models are used in network design to balance cost and performance. The degree distribution P(k) of a network often follows a power law in real networks, leading to scale-free models that explain the resilience of the Internet backbone.
拓扑模型用于网络设计，以平衡成本与性能。真实网络中节点的度分布 P(k) 通常遵循幂律，由此产生的无标度模型解释了互联网骨干网的韧性。
3. Queuing Theory for Network Performance | 排队论与网络性能
Queuing theory models network nodes as service stations with arrival and service processes. The classic M/M/1 queue assumes Poisson arrivals with rate λ and exponential service times with rate μ. The utilisation factor is ρ = λ / μ, and the average number of packets in the system L is given by L = ρ / (1 - ρ), provided ρ < 1 for stability.
排队论将网络节点建模为具有到达过程和服务过程的服务站。经典的 M/M/1 队列假设到达率为 λ 的泊松到达和速率为 μ 的指数服务时间。利用率因子为 ρ = λ / μ，系统中的平均数据包数 L 为 L = ρ / (1 - ρ)，前提是 ρ < 1 以保证稳定性。
The average waiting time W and residence time T follow from Little's Law: L = λ W. These formulas allow network engineers to estimate router buffer requirements and end-to-end delay, ensuring quality of service (QoS) under expected traffic loads.
平均等待时间 W 和逗留时间 T 可由利特尔定律 L = λ W 得出。这些公式使网络工程师能够估算路由器缓冲需求和端到端延迟，从而在预期流量负载下确保服务质量（QoS）。
4. Traffic Modelling: Poisson and Markov Processes | 流量建模：泊松与马尔可夫过程
Network traffic is often modelled as a stochastic process. The Poisson process is widely used for connection arrivals, while Markov chains describe state transitions such as idle, transmitting, and congested. A continuous-time Markov chain can represent the number of active connections in a link, with birth rates and death rates defining the transitions between states.
网络流量通常被建模为随机过程。泊松过程广泛用于连接到达建模，而马尔可夫链描述空闲、传输和拥塞等状态转换。连续时间马尔可夫链可以表示链路中活动连接的数量，其中出生速率和死亡速率定义了状态之间的转移。
These models support traffic engineering. For example, the Erlang B formula calculates the blocking probability in a circuit-switched network with C channels and offered traffic A:
这些模型支持流量工程。例如，爱尔兰 B 公式计算具有 C 条信道和输入话务量 A 的电路交换网络中的阻塞概率：
Pblock = (Aᶜ / C!) / (Σₖ₌₀ᶜ Aᵏ / k!)
This formula helps dimension telephone trunk lines and cellular base station channels to meet call dropping targets.
此公式有助于确定电话中继线路和蜂窝基站信道容量，以满足呼叫丢弃率指标。
5. Network Flow Models | 网络流模型
Maximum flow and minimum cut models are fundamental to network capacity analysis. Given a directed graph with edge capacities, the max-flow min-cut theorem states that the maximum amount of flow from source s to sink t equals the capacity of the minimum cut separating s and t. Algorithms such as Ford-Fulkerson and Edmonds-Karp compute the optimal flow.
最大流和最小割模型是网络容量分析的基础。给定具有边容量的有向图，最大流-最小割定理表明，从源点 s 到汇点 t 的最大流量等于分离 s 和 t 的最小割容量。Ford-Fulkerson 和 Edmonds-Karp 等算法可计算最优流。
In practice, these models are applied to WAN bandwidth provisioning, data centre load balancing, and content delivery network (CDN) traffic routing. The generalised multi-commodity flow problem further enables fair bandwidth allocation among competing flows.
在实践中，这些模型应用于广域网带宽规划、数据中心负载均衡以及内容分发网络（CDN）流量路由。广义的多商品流问题还能够在竞争流之间实现公平的带宽分配。
6. Petri Nets for Protocol Modelling | 用于协议建模的Petri网
Petri nets are a graphical and mathematical modelling tool for describing concurrent, asynchronous, and distributed systems. A Petri net consists of places, transitions, and tokens. The state of the network is represented by the distribution of tokens among places, and transitions fire according to enabled rules, capturing protocol handshakes and resource sharing.
Petri 网是一种用于描述并发、异步和分布式系统的图形化数学建模工具。Petri 网由库所、变迁和托肯组成。网络状态由托肯在库所中的分布表示，变迁根据使能规则触发，从而刻画协议握手和资源共享。
Petri nets are used to verify correctness of network protocols such as TCP, where states like SYN-SENT, ESTABLISHED, and FIN-WAIT are modelled. Reachability analysis and invariant checking can detect deadlocks and buffer overflow conditions before implementation.
Petri 网用于验证网络协议的正确性，例如对 TCP 中 SYN-SENT、ESTABLISHED、FIN-WAIT 等状态进行建模。可达性分析和不变量检查能够在实现之前检测死锁和缓冲区溢出情况。
7. Applying Models: Routing Algorithms | 模型应用：路由算法
Routing algorithms use mathematical models to find optimal paths. Dijkstra's algorithm computes shortest paths in weighted graphs, forming the basis of OSPF (Open Shortest Path First). The Bellman-Ford algorithm supports distance-vector protocols such as RIP by iteratively updating path costs.
路由算法使用数学模型来寻找最优路径。Dijkstra 算法计算加权图中的最短路径，构成 OSPF（开放最短路径优先）的基础。Bellman-Ford 算法通过迭代更新路径开销来支持距离矢量协议（如 RIP）。
Beyond static metrics, dynamic models incorporate link congestion and delay as cost functions. Game-theoretic routing models treat flows as strategic players, leading to stable equilibria and more resilient traffic distribution in software-defined networks (SDN).
除了静态度量外，动态模型将链路拥塞和延迟作为代价函数。博弈论路由模型将流量视为策略性参与者，从而在软件定义网络（SDN）中实现稳定均衡和更具韧性的流量分配。
8. Network Reliability Models | 网络可靠性模型
Reliability is the probability that a network remains operational under component failures. For a series system of n independent components with reliabilities rᵢ, the overall reliability is Rseries = ∏ᵢ rᵢ. For a parallel system, R_parallel = 1 - ∏ᵢ (1 - rᵢ). These basic models extend to bridge and mesh structures using decomposition and Monte Carlo simulation.
可靠性是网络在组件故障下仍保持运行的概率。对于由 n 个独立组件组成的串联系统，其可靠性为 R_串联 = ∏ᵢ rᵢ。对于并联系统，R_并联 = 1 - ∏ᵢ (1 - rᵢ)。这些基本模型可扩展至桥式和网状结构，使用分解法和蒙特卡洛模拟。
Reliability models guide redundancy design and Service Level Agreement (SLA) guarantees. By analysing failure rates, mean time between failures (MTBF), and repair rates, operators can calculate availability A = MTBF / (MTBF + MTTR), where MTTR is the mean time to repair.
可靠性模型指导冗余设计和服务水平协议（SLA）保证。通过分析故障率、平均故障间隔（MTBF）和修复率，运维人员可以计算可用性 A = MTBF / (MTBF + MTTR)，其中 MTTR 为平均修复时间。
9. Simulation and Numerical Analysis | 仿真与数值分析
When analytical models become intractable, discrete-event simulation is used. Network simulators such as ns-3 and OMNeT++ implement statistical models and generate performance metrics like throughput, latency, and packet loss. Random number generators drive source traffic based on empirical distributions, enabling realistic evaluation.
当解析模型变得难以处理时，可使用离散事件仿真。ns-3 和 OMNeT++ 等网络模拟器实现统计模型并生成吞吐量、延迟和丢包等性能指标。随机数生成器基于经验分布驱动源流量，从而实现逼真的评估。
Numerical analysis also plays a role in solving fixed-point equations for congestion control, e.g., TCP throughput as a function of loss probability p and round-trip time RTT. The approximate throughput T = 1.22 / (RTT · √p) helps engineers tune protocol parameters without full-scale tests.
数值分析还在求解拥塞控制的定点方程中发挥作用，例如将 TCP 吞吐量表示为丢包概率 p 和往返时间 RTT 的函数。近似吞吐量 T = 1.22 / (RTT · √p) 帮助工程师无需开展全面测试即可调整协议参数。
10. Limitations and Future Trends | 局限性与未来趋势
Classical models often assume stationarity and independence, which do not always hold for real traffic showing self-similarity and long-range dependence. More recent models, such as fractional Brownian motion and deep-learning-based predictive models, address these challenges by capturing burstiness and temporal correlations.
经典模型通常假设平稳性和独立性，而真实流量具有自相似性和长程依赖性，这些假设并不总是成立。较新的模型，如分数布朗运动和基于深度学习的预测模型，通过捕捉突发性和时间相关性来应对这些挑战。
Future network architectures—including 5G/6G, the Internet of Things (IoT), and quantum networks—require hybrid models that integrate queuing, graph, and control theories. Mathematical models will remain essential for ensuring scalable, secure, and energy-efficient communication.
未来的网络架构——包括 5G/6G、物联网（IoT）和量子网络——需要融合排队论、图论和控制理论的混合模型。数学模型仍将是确保可扩展、安全和节能通信的关键。
11. Conclusion | 总结
Mathematical models transform networking from an empirical craft into a rigorous engineering discipline. From graph theory and queuing analysis to Petri nets and reliability formulas, these tools enable precise description, prediction, and optimisation of network behaviour. Mastery of these models is essential for any computer science student preparing for examinations and practical network design.
数学模型将网络从经验技艺转变为严谨的工程学科。从图论和排队分析到 Petri 网和可靠性公式，这些工具能够精确地描述、预测和优化网络行为。对于任何准备考试和从事实际网络设计的计算机科学学生来说，掌握这些模型至关重要。
Published by TutorHao | Computer Science Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图