On 8 May 2000 michael@dgmo.org wrote:
> Linus Torvalds <torvalds@transmeta.com> writes:
> [ ... ]
> > Happily, the "multiple mmaped writers" case is extremely rare and usually
> > doesn't make any sense at all (which is probably why it is rare), so the
> > performance issue is probably not really a very realistic one.
> >
> > [ Famous last words. I wonder when the next mindcraft comes along and
> > creates a benchmark for this.. It would have to be fairly contrieved, I
> > suspect. ]
> Unfortunately, no. :-(
>
> Once apon a time there used to be a news server that used
> a mmap()'ed active and history file with lots of process
> receiving news and writing updates into the active and
> history files. Alas, poor thru-put. I knew it,
> Horatio....
Ever more so with HighWind's Typhoon product. Their news server product
is highly multi-threaded, and makes extremely heavy use of mmap(),
including a lot of parallel mmap writers. Their software even broke
some early 2.2.x kernels -- exposed some threading bugs IIRC.
Jeff
-
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/
This archive was generated by hypermail 2b29 : Mon May 15 2000 - 21:00:12 EST