Why does disk I/O slow down a CPU bound task?

Dave Johansen davejohansen at gmail.com
Tue Mar 31 01:10:24 UTC 2015

On Mon, Mar 30, 2015 at 3:15 PM, M. Edward (Ed) Borasky <znmeb at znmeb.net>

> On Mon, Mar 30, 2015 at 2:43 PM, drago01 <drago01 at gmail.com> wrote:
> >
> > https://lwn.net/Articles/572911/
> > https://lwn.net/Articles/467328/

Thanks for the info. That's very helpful, but it looks like that discussion
didn't come to any sort of resolution. Do you know if it's being picked up
anywhere else?

Yeah, I vaguely remember that discussion. I went down this rabbit hole
> in mid-2008, wrote a couple of papers and moved on with my life.
> There's no software solution - get faster hardware and more RAM and
> don't design I/O-bound systems.

That works for a lot of problems, but some are I/O-bound by definition.
Databases are a prime example of a task of that (and the one that started
me looking into this). For any non-trivial application, you'll never be
able to throw enough CPU/RAM at the problem to make the disk not matter
when the system in under heavy use.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.fedoraproject.org/pipermail/devel/attachments/20150330/a0e2bce9/attachment.html>

More information about the devel mailing list