I`ve built a Rpeorting structure with jasperintelligence.
Our fact table has round about 150.000 rows, 9mb + 80mb indexes We have the following dimensions:
time ( built like foodmart)120 rows
accounts (parent/child-Hierarchy with closure) 17000 rows, 3 mb + indexes
brands (normal structure) 10 rows
Cost-center (parent/child with closure) 35 rows
subsidiaries (parent/child with closure)80 rows
On all important keys indexes are set in fact-tables and dimension-tables.
Until now i have not considered using aggregatet tables, because i think our datasize is no so big that i`have to do this.
What is the recommended size for RAM and what kind of Processor would you use.
Thanks for your help
Our fact table has round about 150.000 rows, 9mb + 80mb indexes We have the following dimensions:
time ( built like foodmart)120 rows
accounts (parent/child-Hierarchy with closure) 17000 rows, 3 mb + indexes
brands (normal structure) 10 rows
Cost-center (parent/child with closure) 35 rows
subsidiaries (parent/child with closure)80 rows
On all important keys indexes are set in fact-tables and dimension-tables.
Until now i have not considered using aggregatet tables, because i think our datasize is no so big that i`have to do this.
What is the recommended size for RAM and what kind of Processor would you use.
Thanks for your help
3 Answers:
Posted on October 23, 2006 at 12:17pm
This size of data should be fine, but performance can be effected, even at low data volumes, but what you are doing.
Are you having performance issues now? Looking at it, the accounts hierarchy may be an issue, even with the closure table.
Under the Admin menu, you can set Tracing = 1 to get logging of SQL being executed and timings.
Sherman
JasperSoft
Are you having performance issues now? Looking at it, the accounts hierarchy may be an issue, even with the closure table.
Under the Admin menu, you can set Tracing = 1 to get logging of SQL being executed and timings.
Sherman
JasperSoft
Posted on October 23, 2006 at 12:52pm
Yes the performance is not as good as i supposed!
a queries takes upto 2 min bevore the data is calculated. and that happens when data is already in the cube(the java.web.exe, size 132mb)
where can i set the tracelevel? this can not be done through the repository in the JI web-application directly?
a queries takes upto 2 min bevore the data is calculated. and that happens when data is already in the cube(the java.web.exe, size 132mb)
where can i set the tracelevel? this can not be done through the repository in the JI web-application directly?
Posted on October 23, 2006 at 1:26pm
Sorry, I got mixed up between the JI open source project versus the commercial product. In the OS project, you can view OLAP properties, but not change them interactively. The settings are in /WEB-INF/mondrian.properties.
Here is what you can add:
Here is what you can add:
| Code: |
<br /> ###############################################################################<br /> # Property which controls the amount of tracing displayed.<br /> # <br /> # If trace level is above 0, SQL tracing will be enabled and logged as<br /> # per the "mondrian.debug.out.file" below. This is separate from Log4j<br /> # logging.<br /> mondrian.trace.level=1<br /> <br /> ###############################################################################<br /> # Property containing the name of the file to which tracing is to be<br /> # written. If empty (the default), prints to stdout.<br /> #mondrian.debug.out.file=<br /> <br /> </td></tr></tbody></table><br /> <br /> In our commercial releases, we have added the ability to change these properties through the UI and also have performance monitoring with reporting to see which queries are slow etc.<br /> <br /> <br /> Sherman<br /> JasperSoft |
