星载大容量固态存储器快速可靠启动算法设计

Design of quick initialization algorithm for space-borne solid state recorder

为解决星载SSR启动缓慢,可靠性不足所带来的无法满足复杂、灵活的任务需求问题,研究星载大容量固态存储器(SSR)的索引建立机制,分析传统方案的优缺点并结合在轨运控需求、工作模式等特点,提出一种适合基于NAND FLASH的星载大容量固态存储器的快速启动算法.原有文件系统存储索引表来自于NAND FLASH空余区,算法增加新的保留区设计及相应的启动过程,避免重启时扫描空余区重新建立索引.针对空间环境单粒子效应带来的存储错误,NAND FLASH保留区的索引信息采取ECC编码、冗余备份、分区存储等可靠性措施,提高索引表的可靠性.本文介绍了应用保留区的启动工作机理以及不同模式下的更新方式,阐述了系统在功能性重启和故障性重启下的扫描方式,并建立数学模型分析算法的有效性,最后在使用欧比特NAND FLASH搭建的测试平台上进行验证.算法功能性重启索引建立耗时5.643 ms,故障重启索引建立73.985 ms,而传统算法重启索引建立50.37 s.实验结果表明,本算法显著减少了系统启动耗时.

The starting up of onboard SSR is slow and the management is unreliable,they can not satisfy the demand of the mission which is becoming more and more complex and flexible. To solve the problem,this paper focuses on the index establishing mechanism for onboard Solid State Recorder（ SSR）,a quick initialization method for onboard massive capacity SSR based on NAND FLASH is proposed by analyzing the advantages and disadvantages of traditional scheme,the requirement of on-orbit operation and control,and working mode,etc. The proposed method uses the design of reserved area and corresponding booting process on the basis of the original NAND FLASH indexing mechanism by space area,which largely reduced the organization time for index table. By considering the memory error caused by single particle effect,the index information in the reserved area of NAND FLASH uses reliability methods including ECC coding,redundancies and partition storage to improve the reliability of the index table. The restart working mechanism and the update of different mode for the reserved area of application is firstly introduced. Followed by the description of the scanning mode in both functional restart and fault restart,and a mathematical model is built to prove the effectiveness,and is verified in a platform established using Orbita NAND FLASH. Finally,in the designed platform the functional restart of the algorithm takes 3.643 ms and the fault restart takes 73.985 ms while the traditional algorithm takes 50.37 s,which demonstrate that the proposed algorithm significantly reduces the mounting up time.