Re: [2.4.17/18pre] VM and swap - it's really unusable

[Marcelo Tosatti]

On 21 Jan 2002, Robert Love wrote:

> On Mon, 2002-01-21 at 16:49, [e-mail] wrote:
> 
> > > (average of 4 runs of `dbench 16')
> > > 2.5.3-pre1:		25.7608 MB/s
> > > 2.5.3-pre1-preempt:	32.341 MB/s
> > > 
> > > (old, average of 4 runs of `dbench 16')
> > > 2.5.2-pre11:		24.5364 MB/s
> > > 2.5.2-pre11-preempt:	27.5192 MB/s
> 
> > Robert, with all due respect, my tests of dbench show such high
> > variation that 4 miserable runs prove exactly nothing.
> 
> Well you asked for dbench.  Would you prefer 10 runs each?  There were,
> however, no statistical anomalies and the variation was low enough such
> that I suspect I could construct a reasonable confidence interval from
> these 16 runs.
> 
> I've run these tests over and over again sufficiently that the
> repeatability of obtaining improved marks under a preemptive kernel is
> evident to me.
> 
> You can see very old (2.4.6) yet still positive results from Nigel, too:
> http://kpreempt.sourceforge.net.
> 
> I guess the point is, everyone argues preemption is detrimental to
> throughput.  I'm not going to argue that we aren't adding complexity,
> because clearly we are.  But now we have tests showing throughput is
> improved and people still argue.  I've seen the same behavior under
> bonnie, timing kernel compiles, etc ...

Sure, you've seen it. But _why_ it happens ?

That is the point.