基于多Agent的实时数据库排错研究

针对容错技术对错误处理的不彻底现状,提出从混沌信息流中自动识别并排错的新思路,彻底根除错误,完成故障的“预防“工作.排错与容错前提下的排障有着根本区别,排障是从叶子出发局部而滞后处理,排错新思路是从信号的根本特征出发,分析错误信息与有效信息之间的关系,对错误进行定义,最后提出排错的解决方案,分两步将信息流中的错误分离出来.具体实现结合多Agent技术和桶技术自适应地产生常态特征标准,用常态特征标准将混沌信息一分为二,划分出非常态信息;再进行逻辑诊断,将非常态信息又一分为二,确诊错误.研究从根本上排除与消除混沌信息流中的错误,将大规模计算转化为小规模可行的计算问题.

New idea of fault-removing-RDB by multi-agent

A new idea that faults removed and identified automatically from chaotic information flow was proposed in order remove faults thoroughly and takes precautions against them in advance.There is a fundamental difference between fault-removing and error-removal on the premise of fault-tolerance.Error-removal started from leaves to local control behind error.But new idea of fault-removing starts from root of fundamental character of signal to analyze the relation between available information and fault information.The conception of fault was defined and a new solution of fault-removing was improved.The fault can be separated from chaotic information flow by two steps: applying agent technology to forming normal-criterion adaptively and using this normal-criterion to divide the chaotic information into two parts and discriminate the non-normal information from normal-criterion information;diagnosing the non-normal information thought logic judgments to identify the faults.How to remove a fault from chaotic information flow was presented and transfer the large-scale calculations were turned into small-scale practicable calculation problem.