RE: Max tcp connections

David Schwartz (davids@webmaster.com)
Fri, 12 Nov 1999 00:12:13 -0800

> David Schwartz wrote:
> > > Signal based I/O is definitely the right approach for scaling
> >
> > I still find this very hard to believe. How can one
> fast pass through poll
> > (assuming a smarter implementation) be less efficient than the
> OS sending me
> > 5,000 signals? That strikes me as an incredible claim.
>
> Easy. The kernel doesn't send you 5,000 signals. If the
> signal queue starts overflowing, it gives up and sends you
> a single SIGIO, and you call poll().

That's very cool. If you call poll with a timeout of zero, is it smart
enough not to put you on any wait queues? That would be the best of all
worlds.

> The signal model gives you the best of both worlds: low overhead
> when only a few fd's at a time need attention (no linear scan),
> and low overhead when they all do (same as old poll() method,
> but with the extra overhead of a single SIGIO).
>
> sct must be sleeping, or he would have answered like this, I think :-)

That's pretty cool, actually.

DS

-
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/