Tuesday, March 20, 2012

Could RS's data mining be used in a production setting?

I just have a simple question that can hopefully answer whether I should dive too deeply into Reporting/Analysis Services. I'm looking to run some clustering algorithms on user-inputted keywords on objects within my database (please see http://flickr.com/photos/tags/friends/clusters/ for an example of what I'm talking about). It seems to me that the various reporting methods that SQL Server provides would be an ideal, quick and easy way of providing this for my SQL Server 2005 database. I envision having some sort of job that would get the data from the reporting service daily or hourly and insert it into my main database.

Can the reporting services provide this type of functionality? And if so, would this be scalable? I would want to be able to access this clustered data in much the same way I do queries across my database and would want them to be done quickly and efficiently.

Thank you for your help.Can you clarify what you mean by "get the data from the reporting service"? In general, you could have a mining model and produce a report from it or you could run a report that creates a temporary (session) mining model.|||I'm not entirely sure on the concept myself. I was thinking that accessing the mined data from whatever form reporting service uses on a production site would be too slow, wouldn't scale. Rather, I'd like to migrate that data and transform it on some schedule to the main SQL Server database so that the queries would be quicker and scalable.|||

When you process a data mining structure, the discovered patterns are saved in an SSAS database (the same storage format as with a cube) so the data mining prediction queries should be fast. Nevertheless, I would suggest a POC to prove that your scalability requirements are met.

sql

No comments:

Post a Comment