performance differnce between 3027 and 3033

Hi,

I have found major differnces in performance between the two versions - specially with queries that should be answerd from cache if i execute multiple times...

Test Query - after the first execute it tooks

1 second with 9.0.3027 and 12 seconds at 9.0.3033.

Is this a bug in the last ctp - does this change back in the released SP2? or is the something bad in my models for the latest version

we used two equal machines - processed the db on 9.0.3033 and restored the backup on 9.0.3027

query looks like

select

{[4WochenSet]* [Zeit].[Tag].[Tag].Members} on axis(0),

Descendants([Immobilien].[Geographie].[Bundesland].&[Wien]&[AUT], [Immobilien].[Geographie].[PLZ],SELF) on axis(1)

from

Angebotsanalyse

where ([Measures].[Kaufpreis je Wohnflaeche MW], [Plattform].[Plattform].&[156],[Immobilien].[Immobilientyp].&[5])

LG, HANNES

Are there any calculations used by the query? How is the set [4WochenSet] defined? If you capture the performance counters under MSAS 2005:MDX during the query, do you see any significant difference in performance counter values between the two builds? For example, do you see a much larger increase in the value of the [Total calculation covers] performance counter in build 3033? If there is calculation involved in the query, such IIF, try connection string property "Cache Policy=9" to see if there is any difference in performance.

Jeffrey

|||

There are some calc, the measure [Kaufpreis je Wohnflaeche MW] is defined as

AVG(existing({[Immobilien].[Immobilie].[Immobilie].Members}),[Measures].[Kaufpreis je Wohnflaeche])

and the set is defined as

{Tail(except(

exists({[Zeit].[Zeit nach Wochen].[Woche].Members},[Zeit].[Abgeschlossen].&[True]),

exists({[Zeit].[Zeit nach Wochen].[Woche].Members},[Zeit].[Abgeschlossen].&[False])

),4)}

by this definition there are no if calculations, but the base measure [Measures].[Kaufpreis je Wohnflaeche] is defined as lastnonempty.

At the moment i test only using management studio - later i will test it with your sugestions.

Is this further optimized in the released version of SP2? so we do not need the setting?

LG, HANNES

|||

It's hard to tell what caused the performance degredation. But since there is no IIF and other calculation conditions, "Cache Policy=9" shouldn't make any difference.