On 24 Jan 2001, David Wragg wrote:
> Marcelo Tosatti <[EMAIL PROTECTED]> writes:
> > Additional caching of block locations makes read/write (once we have it)
> > clustering more efficient because it avoids additional accesses to disk
> > (which are going to cause seek time in the worst case).
>
> Are you suggesting changing the block allocation behaviour basd on
> which bitmaps are in memory?
No. I was refering to read/write clustering (which is something we want
for 2.5). The idea is to build clusters (groups) of physically contiguous
data (pages, in our case) and write or read them together in big IO
operation(s), avoiding disk seeks and minimizing IO overhead.
Now lets say we want to write a dirty page in memory which has offset 10
inside its inode. To know wheter you can cluster the pages at logical
offsets 9 and 11 (for example) you have to check if these pages are
physically contiguous on the underlying block device also.
If you're caching metadata more aggressively (like you're doing with the
block bitmap readahead) than usual, you have a bigger chance of finding
the physical block number for a given page, avoiding disk accesses.
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]