• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2007, Vol. 29 ›› Issue (8): 82-87.

• 论文 • 上一篇    下一篇

空间探测中基于COTS部件的软件容错技术

杨学军 高珑   

  • 出版日期:2007-08-01 发布日期:2010-06-02

  • Online:2007-08-01 Published:2010-06-02

摘要:

随着航天活动的发展,空间探测任务对于高性能计算的需求越来越明显,高性能的空间超级计算机成为决定下一代空间探测计划成败的关键之一。专用的防辐射计算部件不仅价格昂贵,而且在性能上远远落后于同时代的商用部件(COTS)。使用软件容错技术在COTS部件上搭建空间超级计算机,在达到和专用防辐射部件同样的容错效果的前提下 ,能够大幅度降低成本,提高性能和性能/功耗比。美国国家宇航局和斯坦福大学的实验已经验证,使用COTS部件有助于实现低成本高效能的下一代空间科学探测计划。

关键词: COTS部件 空间超级计算机 软件容错 低成本 高效率 高可靠

Abstract:

As man explore into deeper space, the need for onboard supercomputers becomes urgent. High performance onboard supercomputer becomes a key issue in th e next-generatlon space explorations. Radiation-hardened components are not only expensive, but also lag several generations in performance behind COTS components in the market of supercomputing technologies. Onboard supercomputers built on COTS components, hardened with software fault tolerance technol ogies, can have an equal fault tolerance ability of their radiation hardened counterparts, but at a much lower cost, with a much higher performance and  higher performance/power ratio. Experiments in NASA and Stanford University have proved that, software fault tolerance on COTS components can help imple  ment low-cost and efficient next-generation space explorations.

Key words: (COTS, onboard supercomputer, software fault tolerance, low-cost, efficient, reliability)