I think there is a product called "RamDoubler" for the Macintosh
which implements a two-stage virtual memory system:
1) As usual, there are a bunch of uncompressed pages in use.
2) A first-level swapout compresses a page into an area in memory.
I think RamDoubler allocates more than half of memory to holding
compressed versions of swapped-out pages.
3) A second-level swapout writes the (compressed?) page to disk.
If your process has a working set which is a bit too large to fit
in physical memory, but which, compressed, fits in physical memory,
then it may run a bit faster since there is no disk I/O. For lots
of applications, single-process latency is the most important
measure of performance.
My guess is that a compressed swap would be most useful in an
embedded application which has no hard disk and is trying to get
by on the minimum possible RAM. In this situation, you would be
back to a single-level swap system. Does anyone know if Linux is
getting into the embedded world much?
-Iain
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/