Re: Oracle Benchmark Results for Different Hardware Configurations?

HansF wrote:
> On Fri, 15 Sep 2006 11:29:58 -0700, DA Morgan wrote:
>
> > Serge Rielau wrote:
> >> Further, car manufacturers certainly tune there vehicles to get the
> >> safety rating they aspire to rather than decreasing the de-facto injury
> >> rates (which would be the real application).
> >
> > But it is also what they sell. No RDBMS vendor can claim that what I
> > buy from their salesperson is exactly what was tested. A claim that
> > sleazy car salesman can make with a straight face.
> >
>
> What is the purpose of the commercial benchmark?
>
> I believe it is to give a relative comparison so that someone looking for
> equipment can use that as a starting point for decision making. Using
> defined standards to ensure intelligent people have a basis for trusting
> that comparison.
>
>
> I just bought a new car. One of the qualifications was 'very good fuel
> economy'. I looked at the standard fuel rating specs to assist in the
> decision process.

So, what did you get? (I seem to recall a flap among some hybrid buyers
when they found the highway mileage wasn't near what the sticker says).

Personally, I rate NVH over economy, as I drive a lot, then have to think to
get paid. There were a couple fuel bills this summer that almost changed
my priorities, but not quite.

>
> I would be a fool to expect the same mileage as the car manufacturer has
> all sorts of tricks to squeeze the best out of the car.

Yeah, the problem was hybrids have their big advantage in citified driving,
people's expectation was a bit off that.

>
> But I would be just as big a fool to expect the manufacturer or sales
> house to give me 10 cars, for one week each, so I could make my own
> comparison under my normal driving conditions.

Well, they _do_ give car writers cars, some of whom are leadfoots.

>
> So I used the benchmark as a guideline. Used it to reject some cars and
> select others for further study. I'll be happy if the new car achieves
> within 15% of the specs ... I simply don't drive downhill with a tail wind
> all the time.

That's reasonable.

>
>
> Same thing with the TPCs ... I know that the load is not identical to mine
> and that my tuning skills are not as extreme as those by the TPC team.
> But I also realize that all the vendors use the 'professional driver on a
> closed track' tactic, so I know that the benchmarks describe 'the best
> possibility' regardless of vendor combination. Therefore I can look at
> the benchmark results and select or reject configurations //as candidates
> for further study//!

I think the equivalent to the car writers is finding other customers using the software
and hardware combinations. But then again, I think people buying applications
are more inclined to consider funcionality than benchmarks. They often ask
their IT people to give input and _the IT people_ look at benchmarks - and I think that
part is often done wrong.

(One of the car magazines did compare some hybrids to one of the high-mileage
cars - and a Corvette - focusing on mileage. The surprising part was how well the Corvette did, (
http://www.autoweek.com/apps/pbcs.dll/article?AID=/20060424/FREE/60417021&SearchID=73244466514443 compare hybrid Accord - and 3/4 the mileage of the Prius).)

>
> But don't accuse me of being stupid enough to expect the TPCs to reflect
> my workload, or to give me an absolute target for performance in my
> environment. That level of stupidity is reserved for the PHB and
> shareholder crowd.

You, no. But others who are technical and should know better, I have seen.
I suppose that's how they become PHB's...

jg

--
@home.com is bogus.
"PostScript from Adobe Systems is an underappreciated yet superb
general purpose computing language. In its spare time, PostScript
also excels at dirtying up otherwise clean sheets of paper." - Don
Lancaster