The first time I read this it was very helpful to me. Thank you.
Since then I gained some insight about the downside of creating indexes:
if you write into a table (UPDATE
or INSERT
) with one index, you have actually two writing operations in the file system. One for the table data and another one for the index data (and the resorting of it (and - if clustered - the resorting of the table data)). If table and index are located on the same hard disk this costs more time. Thus a table without an index (a heap) , would allow for quicker write operations. (if you had two indexes you would end up with three write operations, and so on)
However, defining two different locations on two different hard disks for index data and table data can decrease/eliminate the problem of increased cost of time. This requires definition of additional file groups with according files on the desired hard disks and definition of table/index location as desired.
Another problem with indexes is their fragmentation over time as data is inserted. REORGANIZE
helps, you must write routines to have it done.
In certain scenarios a heap is more helpful than a table with indexes,
e.g:- If you have lots of rivalling writes but only one nightly read outside business hours for reporting.
Also, a differentiation between clustered and non-clustered indexes is rather important.
Helped me:- What do Clustered and Non clustered index actually mean?