面向地震预警场景的千万量级预警信息推送服务研究

Research on Ten-million-level Early Warning Information Push Service for Earthquake Early Warning Scenarios

摘要:

面向地震预警场景中震中区域海量人口及时获得预警信息的减灾需求,解决目前每秒约数十万级别的信息推送能力瓶颈,研究了一种海量预警信息高并发分层级联广播和区域定向推送技术,实现亚秒级千万量级预警信息的推送能力。为验证方法性能,采用95台高性能服务器搭建预警信息多层级联发布平台,其中45台高性能服务器模拟一千万用户进行实验。通过在CM端采用CPU核心绑定以及Zookeeper集群管理等技术,提升单服务器推送性能和集群线性增加的稳定性,在不考虑实际网络延时的情况下具备了秒级千万量级预警信息推送能力,一千万条预警信息整体推送延时在500ms以内,达到了秒级推送一千万条预警信息的性能要求,并通过区域定向推送实际环境检验进一步验证了系统推送能力。

Abstract:

In earthquake early warning scenarios,the timely dissemination of warning information to the large population in the epicentral region is crucial for mitigating disaster impacts. However,current systems face limitations in pushing hundreds of thousands of messages per second. To address this challenge,we have developed a high-concurrency,hierarchical cascade broadcast,and region-specific push technology capable of delivering tens of millions of early warning messages per second. To validate this technology,we constructed a multi-level cascade release platform comprising 95 high-performance servers,with 45 of these servers simulating 10 million users for experimental purposes. We implemented CPU core binding and Zookeeper cluster management techniques on the Central Management (CM) side to enhance both the individual server push performance and the stability of linear scalability within the cluster. Excluding actual network latency,the system demonstrated the ability to push tens of millions of early warning messages per second. The overall push delay for 10 million messages was maintained within 500 milliseconds,thereby meeting the performance requirement of pushing 10 million messages per second. Additionally,the system's push capability was further validated through practical environmental testing,confirming its region-specific push functionality.