Steven,

For the less knowledgeable among us (that means me for sure) can you please advise the definition of the word "heuristic/heuristics" that you frequently use?

Jim Nelson


>I cite the overhead on the database server to maintain a table in physical order (which is what a clustered index is) on tables that are heavily transactioned. On performance-sensitive tables that are heavily loaded with inserts and updates, the heuristic does not apply.
>
>Across the OLTP --> OLAP spectrum, what you say is true at the OLAP end, and progressively gets weaker as we get to the OLTP end, especially at higher volumes.
>
>Thus "All tables should have a clustered index" is wrong. It's an engineering absurdity.
>
>You seem to think I don't know what I'm talking about. All I'm doing is challenging the one-dimensional heuristics you are inventing or parroting without really thinking about (or properly qualifying) the extent of the problem space to which the statement applies.
>
>**--** Steve
>
>>Steve..
>>
>>I cite page 247 of Inside SQL Server 7. I did'nt want you to think I pulled this out of thin air...
>>
>>With that in mind, the only tables that would not need a clustered index would be ones that would not participate in queries - I suppose. At some level, every table participates in a query - to one degree or another...
>>
>>The advice is that if your table only has 1 index, it should be clustered.
>>
>>I am sure you will find something that disputes this. Just know that I have at least, some authority on my side of the argument.
>>
>>A lot of times Steve, you go onto to disagree, but don't cite authority for your opinion. If it is your opinon, fine, but state that.
>>
>>
>>
>>
>>>>> Every table in SQL Server should have a clustered index
>>>
>>>Er, no <s>.
>>>
>>>I specifically don't want to go there. Just know that I know this to be an incorrect general heuristic.
>>>
>>>**--** Steve