Research And Application Of Zabbix-based Full Life-cycle Monitoring Platform In A Smart Marine Fisheries Environment

With the rapid growth of the software-as-a-service for cloud computing concept,the service-oriented system platforms are widely used in a different real-world scenarios.Nowadays,the scale and complexity of the systems have increased dramatically,bringing higher challenges to system monitoring and design.At the same time,under the vigorous promotion of national marine formalization,the Zhoushan City Bureau of Fisheries has established the Zhoushan City Intelligence Marine and Fishery Platform in order to achieve refined management of fisheries safety.The platform carried key information used by the fishery administrative authorities to understand the real-time production dynamics of the fishery and to ensure the safety of fishing vessels at sea.The complexity of the system and the high timeliness of the data requires higher requirement for the monitoring and the design of the platform.The traditional monitoring system has the problem of delayed alarm notification,which leads to abnormal data or system failure on the platform before the background monitoring system has responded to the abnormality,seriously affecting the analysis of real-time information by fisheries managers in the early warning management and emergency rescue applications of the marine fishing platform.In addition,the traditional monitoring systems for the data collection process based on uniform sampling have led to the redundancy of monitoring data and waste of resources(including network overheads and own load calculation)due to the lack of consideration on the fluctuating characteristics of the indicators being monitored on the sea fishing platform.(1)Concerning the problem of untimely alarm notification in traditional monitoring systems,this dissertation proposes a resource utilization prediction method based on the Kalman-LSTM weighted combination model to achieve the function of prediction and early warning of important load indicators.Firstly,the noise is removed from the collected index data by means of Kalman filtering.Secondly,the filtered values and the filtered residuals are used as the input to the LSTM neural network,and the prediction results are obtained by iterative training.Finally,the prediction values obtained from the training of the LSTM neural network and the filter residuals are weighted and summed by appropriately weighted factors to calculate the final prediction values.Experiments results show that the algorithm can effectively improve the accuracy of fault location and the manageability of the cluster of sea fishing platforms.(2)In order to address the problems of data redundancy in traditional monitoring using equal time intervals for data acquisition,this dissertation proposes an adaptive acquisition method based on multi-latitude optimization.By introducing gain and quality index latitudes,the method combines the above prediction algorithm to calculate the optimal sequence of acquisition time points,and implement an adaptive acquisition strategy for monitoring data according to the optimal acquisition time,minimizing network overhead and computational load while ensuring monitoring quality,and improving the resource utilization of sea fishing platform nodes.Experiments prove that the adaptive acquisition algorithm based on multi-latitude optimization in this dissertation reduces the amount of data collected while ensuring the accuracy of monitoring data.(3)The system is divided into four modules: prediction and warning,data collection,monitoring and alerting,and data display.Firstly,for the prediction and warning module,the Kalman-LSTM weighted combination prediction algorithm is introduced as the algorithm support for the load index prediction and warning function,and the algorithm is integrated into the data prediction module through secondary development of the Zabbix source code.Then,in the data display module,to address the problem of poor web display effect that comes with Zabbix,this dissertation uses Grafana,a professional visualization tool,to carry out secondary development to achieve a more intuitive and better display for monitoring information;Finally,based on the performance and stability requirements of the monitoring system in historical data retrieval and analysis,the monitoring data storage is optimized,so as to improve the performance of monitoring indicators Finally,based on the performance and stability requirements of the monitoring system in historical data retrieval and analysis,the monitoring data storage is optimized to improve the performance of monitoring indicators and log retrieval and analysis of the sea fishing platform.