Hi OLAP experts,



I've got a couple of performance related questions, they are:



We are having very bad READ performance while users are writing back to the
cube. READ performance is OK when no one is writing back to the cube, my
question is,



-If I have someone writing back to the cube, while others are trying to
read, how does AS handle this? Can readers work concurrently with writers?
Or are readers held out until the write operation is completed? What locking
strategy does AS use?



Cube Processing & bad query response time. We process a general ledger cube,
every 2 hours during business hours, and we are seeing performance issues
when the cube finishes processing.



-My question is, if I have users reading and/or writing to a cube while it
is being processed. What happens at the point that the shadow cube has
finished processing and needs to be committed? That is, if I have a user
executing a large query off of the current cube, will the commitment of the
shadow cube be suspended until the current query is completes? In addition,
what happens to queries that are submitted after the shadow cube wants to
commit? From my observations, I believe users are put on hold.



My theory is that, if I have a user running a big query that ties up AS, the
commitment of the new cube is put on-hold, and any new queries are also put
on-hold.



Thanks in advance

John

This is a very comprehensive paper on performance and AS

http://www.microsoft.com/technet/tre...e/AnSvcsPG.asp

HTH

--
Ray Higdon MCSE, MCDBA, CCNA
---
"news.microsoft.com" <john.holton (AT) no_spamPulte (DOT) com> wrote