Can you explain which time we use clustered index & which time we use non-clustered index & why?

Regards
Anil Singh

Can you explain how clustered index work ? & how to fetch the data from the table when table have clustered index ?

Regards,
Ashok

Query 1:
SELECT d.*,e.* FROM Department d ,Employee e ;

This has clustered index scan on Employee and Department

Query 2:

SELECT d.*,e.* FROM Department d ,Employee e where d.dno=e.dno ;

this has clustered index scan on Employee and clustered index seek on department. But the cost is greater than query 1 .

Pleease explain.

regards
sajitha

A CLUSTERED INDEX stores the data of the TABLE itself according to the INDEX.

A non-CLUSTERED INDEX points to the CLUSTEREd INDEX.

IIUC, the speed increase for a CLUSTERED INDEX is when multiple records are returned that are consecutive in the INDEX. If the INDEX is CLUSTERED, the retrieval will be faster because it only has to check one area of the file, possibly one disk seek. A non-CLUSTERED INDEX is likely to have the data all over the data file, requiring more time for data retrieval.

by using index, how it is going to fastren the retrieving data from a table?

in noncluster index, it does not stores the column data on which we created nc index.

my question is, how it is going to retrieve the data fast when we use nc index column in select and where condition?

i have little bit confusion on this topic. Please anybody clear my confusion. i am very greateful to u in this regard.

thank you,
vego

1. When you create a non-clust on a heap, it creates a B-tree with each node containing the [ non-clust keys + rowid ( i.e. the location of the data record )].

2. When you create a clust and a non-clust index on a table, it again creates a B-tree but with each node having [non-clust Key + clust key ] . Hence it did a seek, also can be termed as covering index in this case.

When you say non-clust index points to the clustered index that means its a bookmark lookup ( in 2000 terminology ).

as per your article … IMHO

observation: non-clustered index seek when a clustered index present, else a table scan

conclusion: when clustered index is introduced, it forces non-clustered index to map itself to Clustered index (internally) hence it’s seek, else it is independent.

assumption: some relation has to be present between Clustered and Non-Clustered index.

I am not sure if the conclusion is correct based on assumption. I would love to test, but dont have the time right now …… We need to take into account the cost of a table scan, non-clustered index seek etc … by the Query optimizer. Selectivity of an index generally plays a role also.

Has anybody tried this test on a table with say a 1000 rows instead of 5 ???? with almost unique data in last two columns ??

Interesting part of above execution plan is now both queries are using nonclustered index SEEK (in place of scan)

When I add the clustered index, both queries are an index seek.

According to your documentation, both queries do an index seek with the NC index and without the CL index.