嵌入式文件系统缓存管理机制研究

Research on Cache Management in Embedded File System

机载计算机对文件系统的性能提出了很高的要求,其缓存机制虽能有效提升读写性能,但在突然断电时可能丢失关键数据。针对该问题,提出了大数据缓存禁止与多模式缓存回写机制:单次的大数据IO请求不作缓存处理,直接写穿到存储设备;记录信息关键但数据量小时采用修改触发模式,一旦执行文件写入操作,立刻会将该数据直接写入存储设备;记录信息数据量大时采用周期触发模式,文件系统的缓存利用率超过一定比例,将缓存脏数据强制回写到存储设备。此外,文件系统每间隔一段时间将缓存脏数据强制回写到存储设备。通过这种混合缓存管理机制,能够在改善嵌入式文件系统性能的同时提升其可靠性。

The runtime environment of airborne embedded file system is bad,caching mechanism can effectively improve read and write performance,but critical data may be lost when a sudden power failure occurs.In view of this problem,this paper puts forward large-data cache disable and mixed mode cache write-back mechanism.Single large-data IO request is not cached.When the recorded information is critical but the amount of data is small,using modified trigger mode.Once the file writing operation is performed,the data will be written directly to the storage device.When the amount of data recorded is large,using periodic trigger mode.When the cache rate of the file system exceeds a certain percentage,it mandatory writes back of cached dirty data to storage device.In addition,the file system mandatory writes back of cached dirty data to storage device at interval.The hybrid cache mechanism can improve the performance while promoting the reliability of the file system.