 | |
Single Source table from Oracle
microsoft.public.sqlserver.olap microsoft.public.sqlserver.olap
 |
| | Thread Tools | Display Modes |
#1
 
| |||
| |||
|
Single Source table from Oracle -
10-16-2006
, 05:01 PM
#2
| |||
| |||
|
|
I have a project in which the customer wants to source all data for a MSAS 2005 cube from an Oracle Database. On top of that, they have put everything in one table, this equates to 25 (yes 25) dimensions. My inital thought (before I found out the data needs to stay in Oracle) was to group the attributes by likeness and treat each grouping like a "junk" dimension then use surrogate keys to do lookups against the groups of attributes. However, I dont see how this is possible if all of the data resides in Oracle. There is no relationships between these attributes, with the exception of the date dim. With that stated, my thought is to build views (25 dims and 1 fact) in Oracle and bring them over into the UDM. Afterwards, create 25 dimensions. The flipside is that these dimensions are really small, with the non-calendar dimension being less than 100 members. Without importing the data into SQL Server, is this the most optimal method? If I was able to load it into SQL Server, is grouping the data into like dimensions (bear in mind that the data is not related, thus large cartesians and large dimensions) the best option or creating 25 dimensions better. Thanks for any insight. |
#3
| |||
| |||
|
|
Hi, You can use one table for both fact and dimensions in SSAS2005. Just Make a cube with the Wizard and choice the table in Oracle. THan you can add all fields from the table as measure or as dimension attribute. If there are hierarcies based on the attributes you can add these as well. Finish with adding calculations and/or KPIs and you have a great cube ! You can use ReportPortal as server/client for using SSAS 2005 OLAP + KPI reports on intranet. www.rpeortportal.com > free 60days trial. Regards, MArco jgb23 schreef: I have a project in which the customer wants to source all data for a MSAS 2005 cube from an Oracle Database. On top of that, they have put everything in one table, this equates to 25 (yes 25) dimensions. My inital thought (before I found out the data needs to stay in Oracle) was to group the attributes by likeness and treat each grouping like a "junk" dimension then use surrogate keys to do lookups against the groups of attributes. However, I dont see how this is possible if all of the data resides in Oracle. There is no relationships between these attributes, with the exception of the date dim. With that stated, my thought is to build views (25 dims and 1 fact) in Oracle and bring them over into the UDM. Afterwards, create 25 dimensions. The flipside is that these dimensions are really small, with the non-calendar dimension being less than 100 members. Without importing the data into SQL Server, is this the most optimal method? If I was able to load it into SQL Server, is grouping the data into like dimensions (bear in mind that the data is not related, thus large cartesians and large dimensions) the best option or creating 25 dimensions better. Thanks for any insight. |
#4
| |||
| |||
|
|
Marco, Thanks for your reply. So based on your response, I would have approximately 25 physical dimensions, based on the single level attributes in my fact table. Is that correct? Thanks "Marco" wrote: Hi, You can use one table for both fact and dimensions in SSAS2005. Just Make a cube with the Wizard and choice the table in Oracle. THan you can add all fields from the table as measure or as dimension attribute. If there are hierarcies based on the attributes you can add these as well. Finish with adding calculations and/or KPIs and you have a great cube ! You can use ReportPortal as server/client for using SSAS 2005 OLAP + KPI reports on intranet. www.rpeortportal.com > free 60days trial. Regards, MArco jgb23 schreef: I have a project in which the customer wants to source all data for a MSAS 2005 cube from an Oracle Database. On top of that, they have put everything in one table, this equates to 25 (yes 25) dimensions. My inital thought (before I found out the data needs to stay in Oracle) was to group the attributes by likeness and treat each grouping like a "junk" dimension then use surrogate keys to do lookups against the groups of attributes. However, I dont see how this is possible if all of the data resides in Oracle. There is no relationships between these attributes, with the exception of the date dim. With that stated, my thought is to build views (25 dims and 1 fact) in Oracle and bring them over into the UDM. Afterwards, create 25 dimensions. The flipside is that these dimensions are really small, with the non-calendar dimension being less than 100 members. Without importing the data into SQL Server, is this the most optimal method? If I was able to load it into SQL Server, is grouping the data into like dimensions (bear in mind that the data is not related, thus large cartesians and large dimensions) the best option or creating 25 dimensions better. Thanks for any insight. |
|
« Previous Thread
|
Next Thread »
| Thread Tools | |
| Display Modes | |
Linear Mode
| |
Powered by vBulletin Version 3.5.3
Copyright ©2000 - 2012, Jelsoft Enterprises Ltd.
Copyright ©2000 - 2012, Jelsoft Enterprises Ltd.