面向粒子物理实验的高速数据流多级动态汇总技术研究

Multi-level dynamic aggregation of high-speed data streams for particle physics experiments

  • 摘要: 现代粒子物理实验前端电子学的数据通道数和数据率急剧增加,给海量电子学数据的高速可靠汇总带来严峻挑战。针对现有数据读出系统在传输带宽和可扩展性方面的不足,提出一种基于VPX(VME International Trade Association eXpanded)总线架构的粒子物理实验高速数据灵活汇总方法。该方法构建了一个分级、可扩展的数据汇总系统,通过光纤与高速链路将前端数据逐级汇总与重组,最终汇总帧通过交换路由传输到远端服务器。系统采用独特的汇总帧结构与动态重组机制,支持前端探测器类型自适应、数据帧长可变与通道数灵活配置。测试表明,该系统汇总链路数据传输效率可高达97%,两级汇总误码率优于3.080×10−15,总输出数据传输速率高达40 Gbps,并能实现汇总级数与通道数的自由扩展,可满足下一代粒子物理实验对大规模前端电子学数据汇总的需求。

     

    Abstract:
    Background Modern particle physics experiments are driving a rapid increase in the channel counts and data rates of front-end electronics, posing severe challenges for the reliable, high-speed aggregation of massive volumes of detector data. Existing data readout systems based on conventional bus architectures are limited in transmission bandwidth and scalability, rendering them insufficient for next-generation experiments.
    Purpose This study aims to overcome these bandwidth and scalability limitations by developing a flexible, high-speed data aggregation method based on the VPX (VME International Trade Association eXpanded) bus architecture, in order to meet the demand of next-generation particle physics experiments for the large-scale aggregation of front-end electronics data.
    Methods A hierarchical and scalable aggregation system was constructed, in which front-end data are progressively aggregated and reorganized through optical fibers and high-speed serial links, and the final aggregated frames are routed to a remote server through a switching network. Primary boards receive data frames from the front-end electronics through bidirectional optical links at up to 1 Gbps per link; on each primary board, a dedicated module performs first-level aggregation and reorganization of multi-channel frames using a purpose-designed aggregation frame structure and a dynamic reorganization mechanism. This design adapts automatically to different detector types and supports variable-length data frames and configurable channel counts. The aggregated frames are then transmitted to secondary boards over high-speed backplane serial links at up to 10 Gbps per link, where they undergo a second level of aggregation before being forwarded over 40 Gbps Ethernet to the remote data-acquisition and storage server. By deploying identical hardware at every level, differentiated only by configurable device IDs, the system allows both the number of aggregation levels and the channel count to be flexibly scaled.
    Results Test results show that the aggregation-link transmission efficiency reaches up to 97%, the bit error rate after two-level aggregation is better than 3.080×1015, and the total output data transmission rate reaches up to 40 Gbps. Moreover, both the number of aggregation levels and the channel count can be freely expanded.
    Conclusions The proposed VPX-based method provides a flexible and scalable framework for high-speed detector data readout. By combining a hierarchical architecture with a dynamic frame reorganization mechanism, it achieves high transmission efficiency and reliability while allowing the aggregation levels and channel counts to be freely expanded, thereby satisfying the demand of next-generation particle physics experiments for the large-scale aggregation of front-end electronics data.

     

/

返回文章
返回