Creating two extracts separately for each table and joining them will give you the same extract as joining the tables and it is not a best practice. As for reducing the extract time, you can use custom sql option to extract the required data only and to make a smaller extract.
Thanks for the reply. I agree that the data structure and data volumes will be same between joining the extract and join the tables. Actually, I do not want to change the data structure and the all the fields are needed.
My understanding is that joining the tables in the view requires the tableau to go back to the underlying tables to join. For my example, it went back through Cisco information server to Oracle while Joining the extract is to join table locally, which is much faster.
1 of 1 people found this helpful
The extract time for both these cases would depend on the specs of the local machine and the server. Also, disadvantage of joining tables locally is not able to schedule refresh to the extract after publishing on to the tableau server and maintaining the extract with the latest data.
Thanks again Shivanakari for the information. It is so true that the refreshing become very weird after using the local extracts separately.
However, I am joining tables in the tableau view and already applied the filters , hide the unused rows , set the aggregation in the month level. However, the importing takes more than 4 hours and still running now. Below is the current rows get retrieved. Does it make sense?
I was told there is 1.7 GB data there and I am wondering whether there is other way for this extract or any of the below action helps?
1. Will create the empty extract and let server instead of the desktop to do work?
2. Can we setup a connection from the data source( oracle server, Cisco information server) to the tableau server directly without going through desktop or occupying local disc?
1 of 1 people found this helpful
Please do not use separate extracts and then join them on the views. This will always be slower than doing it at database side. It will tax your Tableau server as the the data engine is trying to do the job of join. Please do remember that it's the number of columns that will be a dominant factor than the number of rows. Please bring in only required columns that will save you a lot of processing time and hence better performance.
The data seems to be huge at your side. You can consult your DBA, how you can make the query perform better at database aide.
Thanks a lot Suhas. It is very helpful. I have one more question:
We are working on the proof of concept but the business wants to bring a large amount of the data . I am currently saving the extract on my virtual machine( I can not bring it into my laptop desktop because the data source only get connected in the virtual machine) but it is almost the out of space .I am wondering whether publishing the data source to the server and letting the tableau desktop to connect to the server is a good practice or I have to let the company enlarge my VM space and let the extract stay in my local machine.
Please kindly advise Thanks again
You are absolutely right!! Please publish the data source to the sever and connect the desktop to the data source. Howerver, there is a trick involved in making the server do all the work.
There is a great blog ppst http://www.tableau.com/about/blog/2013/9/easy-empty-local-extracts-25152?cb=wgcjbbi6ymz5xw29 ( thanks Kathleen!!) which describes all the steps.
Thanks Suhas for the information. It is very helpful
I tried to make the create the empty extract as the example but when I pull the calculated field into the filter there are three values, yes no and null? Could you please advise why there is null value here ,which does not like the example and will null value has a negative impact here?
Thanks a lot
Create Empty Extract.twbx 1.1 MB