| 注册
首页|期刊导航|天地一体化信息网络|面向天基计算的故障表征与容错机制

面向天基计算的故障表征与容错机制

史冰冰 李泓辛 刘垚圻 周浩 谭海宁 韩银和

天地一体化信息网络2026,Vol.7Issue(1):91-101,11.
✕
天地一体化信息网络2026,Vol.7Issue(1):91-101,11.DOI:10.11959/j.issn.2096-8930.2026010

面向天基计算的故障表征与容错机制

Fault Characterization and Fault-Tolerant Mechanism for Space-Based Computing

史冰冰 1李泓辛 2刘垚圻 1周浩 3谭海宁 4韩银和4

作者信息

  • 1. 中国科学院计算技术研究所处理器芯片全国重点实验室,北京 100089||中国科学院计算技术研究所,北京 100089||上海科技大学信息科学与技术学院,上海 201210||北京中科天算科技有限公司,北京 100094
  • 2. 北京中科天算科技有限公司,北京 100094
  • 3. 中国科学院计算技术研究所处理器芯片全国重点实验室,北京 100089||中国科学院计算技术研究所,北京 100089||北京中科天算科技有限公司,北京 100094
  • 4. 中国科学院计算技术研究所处理器芯片全国重点实验室,北京 100089||中国科学院计算技术研究所,北京 100089
  • 折叠

摘要

Abstract

With the rapid development of commercial aerospace,space-based computing systems are confronted with dual challenges of high failure rates caused by space radiation and the surging demand for computing power resulting from the rapid increase in mission scale.To enhance the reliability and computing efficiency of commercial devices,a four-dimensional system model that includes the physical layer,fault layer,task layer and evaluation layer is constructed,and the"effective average computing power"index to reveal the intrinsic connection between task scheduling and resource allocation is proposed.On this basis,a dynamic fault-tolerant mechanism is designed.Through the coupling modeling of spatio-temporal resources and the adoption of a hybrid optimization strategy,the collab-orative balance of faults-computing power-resources is achieved.Through Monte Carlo simulation evaluation,a hybrid fault-tolerant algorithm integrating redundant configuration,checkpoint optimization and task migration is proposed,which increased the average ef-fective computing power of the system by approximately 39.20%,providing theoretical support and feasible technical paths for the in-orbit reliability guarantee of space-based computing systems.

关键词

容错计算/天基计算/时空资源建模/混合容错算法

Key words

fault-tolerant computing/space-based computing/spatio-temporal resource modeling/hybrid fault-tolerant algorithm

分类

信息技术与安全科学

引用本文复制引用

史冰冰,李泓辛,刘垚圻,周浩,谭海宁,韩银和..面向天基计算的故障表征与容错机制[J].天地一体化信息网络,2026,7(1):91-101,11.

基金项目

国家自然科学基金(No.62025404) (No.62025404)

北京市自然科学基金(No.L241013,No.L258047) The National Natural Science Foundation of China(No.62025404),The Beijing Natural Science Foundation(No.L241013,No.L258047) (No.L241013,No.L258047)

天地一体化信息网络

2096-8930

访问量2
|
下载量0
段落导航相关论文