Re: RFC: design for new VM

From: David Gould (dg@suse.com)
Date: Tue Aug 08 2000 - 02:26:23 EST


On Tue, Aug 08, 2000 at 05:25:36AM -0000, wingel@t1.ctrl-c.liu.se wrote:
> dg@suse.com wrote:
> [snip]
> >and of course VMS is not the
> >last word by any means, but the system was very tunable, and had specific
> >explicit mechanisms to attain many of the goals of vm system.
>
> I'm not sure if I like tunables. My feeling after playing around with
> VMS for a few years is that it's next to impossible to get all these
> tunables right. Having a tunable usually means "we can't get this right
> so we'll just blame the user (admin) for using bad values". Of course
> sometimes tunables are needed, but I'd like as few as possible, or they
> should at least ble automatically managed unless the user sets them
> explicitly.

Me too. I am not sure if I like tunables either. In fact they suck. But
systems that do not work suck even more.

I used to make a living as a VMS tuning consultant, mostly by undoing
whatever the onsite people had been playing with. Later, I got to watch
a few generations of customers (and in house developers) fail to come to
terms with the tuning knobs and levers in the big database servers.

But:

 - we already have a full /proc full of magic tuning
 - as far as I can tell, noone knows how to do it automatically for the
   corner cases, or even just unusual workload mixes
 - and there are lots of high value systems that are tuned just so, and
   will not work with out the tuning.

Bottom line, if we could get rid of tunables, great, but I have no idea
how to do that. And not having them when you need them is much worse
than having them when you don't.

I guess there is a subtheme to my posts today, "the user is often smarter
than the developer thinks".

-dg

> "Just how much can I get away with and still go to heaven?"

nice.

-dg

-- 
David Gould                                                 dg@suse.com
SuSE, Inc.,  580 2cd St. #210,  Oakland, CA 94607          510.628.3380
"I sense a disturbance in the source"  -- Alan Cox

- 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 : Tue Aug 15 2000 - 21:00:14 EST