Data Warehouse Workloads - nonclustered indexes are not required

Hi experts,

i am reading about: Data Warehouse Workloads
but i do not understand:
Large numbers of nonclustered indexes are not required
enrique_aeoAsked:
Who is Participating?

[Webinar] Streamline your web hosting managementRegister Today

x
 
Kent OlsenConnect With a Mentor Data Warehouse Architect / DBACommented:
Hi Enrique,

I've found a couple of articles that use that exact phrase.  The statement is very narrow in scope and refers to specific queries that filter few rows.

In truth, an effective data warehouse requires indexes.  Often times many indexes, depending on the design, data, data volume, and query needs.

Don't get caught up in the misleading statement "Large numbers of nonclustered indexes are not required".  They are required.

The data in a data warehouse is often static (or semi-static).  Update/Insert timing considerations generally don't come into play when determining if the warehouse needs indexes.


Kent
0
 
Jim HornConnect With a Mentor Microsoft SQL Server Developer, Architect, and AuthorCommented:
I'm a developer and not a DBA, so there will be other experts that can give you a more exact answer, but I'll give this a whack...

A table can have only one clustered (i.e. rows are physically ordered) index.  
All other indexes are non-clustered.  Each nonclustered index takes up disk space to store, and processing time when you insert or delete.

The more non-clustered indexes you have, the more memory (bloat) the table stores, and the longer it takes to insert and update.  So there's a certain point of diminishing returns with more indexes.

Also, if tables are huge, and searchability is a big requirement, many companies will make multiple copies of a table for reporting purposes, each one indexed differently (say by customer, address, age, etc) to be optimized for that specific search.

Good luck.
Jim
0
All Courses

From novice to tech pro — start learning today.