i'm just theorizing here, but i'd bet this is another symptom of the
problem i'm chasing with strange buffer cache behavior. it looks to me
like this really isn't a buffer cache problem, but some kind of
pathological interaction between virtual memory management and file system
I/O that limits disk throughput far sooner than it should.
it appears when you start swapping because that's also when shrink_mmap()
is busy stealing pages from everyone in sight, including the buffer cache.
IMHO this is a show-stopping problem for anyone who is serious about using
2.2 for large servers. when the system boots, performance is fabulous,
but over time, it degrades.
- Chuck Lever
-- corporate: <chuckl@netscape.com> personal: <chucklever@netscape.net> or <cel@monkey.org>The Linux Scalability project: http://www.citi.umich.edu/projects/citi-netscape/
- 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/