>If the object is to improve performance by decreasing the amount of data
>transferred across the I/O bus and to disk then you just reserve 4k blocks on
>disk for each page -- the actual amount written may be much less because of
>the compression but you don't have to worry about fragmentation and you still
>have a one to one mapping. This wastes disk space so it doesn't make sense if
>the idea is to reduce the amount of swap needed -- if the idea is to improve
>performance it might work but I guess you would have to know the relative
>speeds of compression/decompression relative to just transferring the stuff
>straight.
This is an interesting concept, but I believe that a major factor in
swapping speed is hard drive seek times. I believe that the main delay in
writing data to disk is seek delay, and writing 1K instead of 4K won't affect
this.
Russell Coker
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu