基于分形细胞和可靠基因的通信网络可靠性模型

The existing reliability models mainly focus on visible physical components (components in software reliability models are usually functions, procedures and modules, they can also be seen as physical ones, because they are actual code lines in different computer languages) and their function failures, and ignore other reliability factors (for example, network configuration and service deployment) and their coupling relationships. This makes them insufficient to model network systems with obvious non-linear characteristics. Based on the idea of systems biology and with communication networks as study cases, the contents of our project include: (1) the identification method of cells for communication networks, characteristics of these cells and coupling relationships of cells; (2) the abstracting method for reliability genes and the their modeling characteristics; (3) a fractal theory and complex system modeling technology based reliability model for communication networks, which has the dynamic evolution characteristics similar to the cell metabolism, proliferation and gene mutation formation; (4) the reliability evaluation method and the regular patterns based on the proposed model. These four contents will initially support a new network reliability theory based on the fractal cell and the reliability gene. The innovations of the project are: a new cell-based model instead of a visible physical components based model; causes of failure are captured as reliability genes in communication network cells, which are more controllable and easier to be modeled for their complex relationships. Meanwhile, the fractal and Multi-Agent System technology are used to describe the dynamic evolution and the emergent properties of complex networks. Our research will provide a new theory and methods for reliability modeling, analysis and optimization for communication networks, which have high requirements of reliability and efficient operation. Furthermore, the theory and methods can also be extended to other network systems in different fields.

针对当前可靠性模型多从可见物理组分入手，主要考虑功能失效而较少考虑其他可靠性影响因素（如网络配置和服务部署等）及相互的耦合关系，难以支持网络系统非线性特征的问题，本项目以通信网络为研究对象，借鉴系统生物学，研究：（1）通信网络细胞识别方法、建模特性和耦合关系；（2）可靠基因抽取方法和建模特性；（3）具有类似细胞代谢、增殖和基因变异等动态演化特征的通信网络可靠性模型；（4）可靠性量化计算方法和可靠性变化规律。初步形成基于分形细胞和可靠基因的可靠性建模新理论。项目特点：以抽象细胞为基础，替代现有模型的可见物理组分；把可靠性影响因素抽象为可靠基因缩小到细胞中，更为可控，能更好支持对多因素及复杂关系建模；采用分形和MAS技术形成类生物体的可靠性模型，实现对网络故障复杂性和涌现性的动态演化支持。本研究能为通信网络提供可靠性建模、分析和设计优化的新理论与新方法，并可扩展应用到其他领域的网络系统。

针对当前可靠性模型多从可见物理组分入手，主要考虑功能失效而较少考虑其他可靠性影响因素及相互的耦合关系，难以支持网络系统非线性特征的问题，本项目从系统论角度出发，以分形技术为基础，研究通信网络细胞和可靠基因概念及其构建方法，为通信网络可靠性建模分析提供一种新视角。具体开展了以下四方面的研究：1）通信网络细胞研究：提出了基于分形重整化技术的网络细胞识别与构建方法，并进一步细分为程序化、随机型细胞（根据业务流程分析构件间的功能逻辑关系，将具有相同业务属性的构件有机组合抽象为网络细胞）；和动态业务细胞（基于业务与构件间动态统计映射关系，将具有同一类映射关系的构件集合抽象为网络细胞）。2）可靠基因研究：以通信网络故障为实际案例，分析总结网络可靠性重要影响因素及相互关系，提出了针对不同网络细胞的可靠基因和抽取方法。3）基于细胞和基因的通信网络可靠性模型：a)基于系统生物学提出了四类网络细胞间耦合关系；b)基因变异研究：提出了基于规则约束和统计数学分布的可靠基因变异方式，作为网络可靠性模型的故障注入；c)基于细胞和基因的通信网络可靠性模型：针对细胞间的耦合关系和业务动态特征，提出了基于网络细胞和基因变异的网络可靠性演化模型。并完成了4）可靠性计算方法与影响分析：提出了a）基于分形细胞的网络连通可靠性评估方法和b)基于程序化细胞的网络拥塞评估方法。研究成果不仅形成了网络可靠性建模与评估新理论，扩展了传统可靠性技术，且相关成果已应用于华为5G承载网（智能光网络），5G核心网（云化虚拟化网络）和5GtoB工业网络，以及机载网络、卫星互联网和车联网等不同领域网络上，解决了企业对网络动态耦合特征难以开展可靠性工作的难题。