On Mon, May 18, 2009 at 11:57:57AM +0200, Stefan de Konink wrote:
On Mon, 18 May 2009, Stefan Manegold wrote:
In case you came to your conclusion from experimental/emperical study, I would be very interested to know about the experiemntes performed and the results measured. In that case, I would be verythankful if you could share the informative details with us, i.e. detailed descriptions of - experimental setup (HW, SW, datasets, workload/queries) - measured results (what was measured how, and what are the results?) - (your) discussion/analysis/interpretation of the results
I thought the MADAM meeting was for this.
Ok. fine. ... we should then revive the ideas of keeping note of MADAMs (e.g., on a Wiki or alike) as reference for those who cannot be present ...
Could you please elaborate on whether you consider you setup/schema "representaive" and/or whether/to which extend you think that your results for one specific case could be generalized to other cases?
Technically the only thing I have to prove to get my code reverted is that there are cases where late contraints hurt performance.
IMHO, our task should be the following: - analyze whether there is indeed a significat performance difference between early and late constraint checking; (5% do not justify any further time to be spent on this) - and if so, analyze under which circumstances which altertative performs (significatly) better or worse than the other; - and (only) then try to find out where the (significat) performance difference comes from - and (only) then + either try to eliminate the difference; + or * make dump/restore use the most suitable alternative (as far as possible with reasonable effort); * and document our experiences to give adivce for users Stefan
Stefan
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |