Currently a filesystem read operation is not optimal because of the following:
Read accept an offset and a length to read
Read then uses the meta to find out what are the file blocks
Using the offset the read opens the needed block file
It does an internal offset in the block file if needed
Then read required length. If the read need to move to the next block file it opens it and continue reading until length is satisfied
This is not very efficient in case of a sequential file read. Let's assume someone opens a large file then start reading it sequentially. This means a single block can be Open, Seeked, then read then closed, then on the next read call the same file will be Open, seeked then read again (specially for large files)
Instead the system need to be improved so feels are not closed immediately (can be kept open in memory in some sort of a cache/map) where they can be closed when are not needed or not used in a while.
Currently a filesystem
readoperation is not optimal because of the following:This is not very efficient in case of a sequential file read. Let's assume someone opens a large file then start reading it sequentially. This means a single block can be Open, Seeked, then read then closed, then on the next read call the same file will be Open, seeked then read again (specially for large files)
Instead the system need to be improved so feels are not closed immediately (can be kept open in memory in some sort of a cache/map) where they can be closed when are not needed or not used in a while.