Question

I like to think I know enought theory, but I have little experience optimizing DB in real world.
I would like to know points of view, thoughts or experiences.

Let’s imagine a scenario like:

Table A
Key: c1, c2, c3, c4
Index: c7, c3, c2

Table B
Key: c1, c2, c3, c4
Index: c1, c5

All are non-clustered.
The tables have 40+ fields.
They are feeded daily by night and have some updates during the day.

Table A, if more queries benefit from Key than Index, could the index impact negatively?
Because insert/delete has to update 2 indexes instead of 1.

Table B, has an extra field at the index, not present in key.

Could a query using c1, c5

Benefit from this Key?:
Key: c1, c2, c3, c4, c5

So that Index could be droped.

What impact does the order of the fields has?
Key: c1, c2, c3
Key: c3, c1, c2

A typical scenario for me is process_date, client_number, operation.
And it feeds with a bunch of data each day (process_date).

Answer 1

if more queries benefit from Key than
Index, could the index impact
negatively? Because insert/delete has
to update 2 indexes instead of 1.

A non-clustered index has a negative impact on insert/update/delete performance. The negative impact is usually outweighed by the increased performance of selects.

Could a query using c1, c5 Benefit
from this Key?: Key: c1, c2, c3, c4,
c5

Yes, if only a few rows share the same c1, then the index will be very effective.

What impact does the order of the
fields has? Key: c1, c2, c3 Key: c3,
c1, c2

Order is important, both for filtering and ordering. An index on (c1,c2) can be used for where c1 = 1 and where c1 = 1 and c2 = 1, but not for where c2 = 1. Likewise, it helps with order by c1, but not with order by c2.