Senior Member
Saiku and jpivot performance
When I run a mdx query in saiku it wil take 2 minutes. When I run the same query in the Schema Workbench it takes less than a second. It happens when I have more than 1 column. When I run the same query for a second time it takes only 2 - 3 seconds.
Does anyone know what the problem could be?
Greetz,
Johan
Senior Member
On the first run a lot of caching is taking place to make the queries faster afterwards.
depending on how much data you are fetching this could take a while. the first run in schema workbench will also be slower.
It also depends on the version of saiku you are using they have made some performance tweaks wich should be in the latest stable release.
Senior Member
I am working with saiku 2.4 and biserver 4.5ce
My fact table has some 10.000 rows.
When I has the dimension with some 1000 memders in the rows and a measure in the columns it takes 1-2 seconds. When I place a dimension with 2 - 5 members in the columns than it takes more then 100 seconds.
A short update
When I use degenerated dimensions. The same query will take 3 seconds in saiku. I find it strange that degenerated dimensions are so much faster.
Last edited by johanhammink; 07-13-2012 at 05:01 AM.
Junior Member
Use index
I suggest you to use index on your db, i suggest B-tree on the the foreign key in the fact_table and do the same on the primaty key in dimensional table, otherwise you can use MonetDB as DBMS, that is column oriented and use automatic index, whit one of this strategies I can ensure that you improve your performance of 80% whitout care about caching, so that means that also the first time that you run a query your performance are good.
I did my research about this, trust me.
G
Originally Posted by johanhammink
Posting Permissions
Reply With Quote