Re: Ideas for abstracting driver IO from bus implementation?

Alan Cox (alan@lxorguk.ukuu.org.uk)
Thu, 18 Mar 1999 21:55:20 +0000 (GMT)


> supposed to be, anyway? Based on this discussion, it sounds like I2O is
> not much more than asymmetric multiprocessing, where you've got special
> I2O processors sitting in various places (maybe on your motherboard,
> maybe on IO boards, maybe some combination) that just run driver code.
> Hasn't this concept been tried and abandoned enough times before?

Yes and no. IBM build supercomputers around it. I2O is also very well
designed (much to my initial suprise). The I/O read and write paths of
the block API are a matter of a couple of hundred instructions, the
controller does all the housekeeping work. It creates very low CPU
loads under high I/O loads. So for some server tasks its tailor made. You
can chain a terabyte of raid storage off your system without spending all
your CPU executing scsi code.

Alan

-
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/