Re: Improving raw network performance [was Apache performance etc.]
Dan Kegel (dank@alumni.caltech.edu)
Tue, 29 Jun 1999 10:23:35 -0700
Jay Thorne wrote:
>
> Dan Kegel wrote:
> > [ cvs change comment for http://www.nl.linux.org/linuxperf/ ]
> > Multiple 100baseT's can clog 2.2 kernel with interrupts. Suggested workarounds.
>
> Which drivers support multiple packets to one interrupt? The tulip chip
> has hardware support for it, do you know if the driver supports it?
>
> Some noise has been made about the NT drivers that switch to polling mode
> under heavy traffic. How much work do you think that will be?
>
> We would need to a) detect heavy interrupt traffic from a driver
> b) tell the interrupt controller to block that interrupt
> c) switch the card to polled mode (assuming the card can
> do it)
> d) start a kernel thread, maybe that polls the card, maybe
> by firing up the timer chip to generate 1000 or 10000 hz timer interrupts,
> instead of 100Hz (for intel,anyway)
> e) run the bottom half with a separate thread, I guess.
That's being discussed on linux-kernel right now, see e.g.
http://kernelnotes.org/lnxlists/linux-kernel/lk_9906_04/msg01523.html
- Dan
-
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/