on 01-13-2016 10:55 AM
We have one full load to cube happens from an InfoObject . Before loading into cube we
1.delete the indexes of cube
2. delete complete cube data along with delting of dimensions
the problem is that it takes 12hrs to load data and it is taking time only at SID generation step nearly 7 to 8 minutes. Total cube size is nearly 4 million. There are no routines, no transformation logic etc..
please provide me solution to increase data load performance to cube.
How far it is a good idea to do number range buffering for dimensions?
NOTE;
i have done the following checks
1.none of the dimension size is not more than 10% of cube size
2.we have increased the dtp job to 6
Hi Praveen
When you completely delete data from a cube and when the data volume is more, it will take time for SID generation.
Even number range buffering wouldn't help much.
Try to do this. This should help for sure.
Instead of deleting the data before loading, just delete index and load the data into the cube.
Once the load is completed, then delete the overlapping request.
By this way, SIDs will not be generated newly and your data will be also be as expected.
Hope this helps
Regards
Karthik
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Praveen,
How your test? If it is good please assign point to Karthik and close this message.
If not, you may try to modify your step 2.
When delete cube data, DO NOT click option to delete the dimensions, then it should do the trick as well.
And you can schedule a job to call RSDRD_DIM_REMOVE_UNUSED to remove unused DIM entry.
(Once a week at weekend for example)
Regards
Bill
User | Count |
---|---|
93 | |
10 | |
10 | |
9 | |
9 | |
7 | |
6 | |
5 | |
5 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.