Have a look at IDENT_CURRENT, @@IDENTITY and SCOPE_IDENTITY() in Books-Online

Win

The tablock lasts for < 1 sec on a 60 million row table, so not a problem for a warehouse database.  If you leave the re-seed job to a mignight...

Analogous to your convention of user-defined functions we prefix our trigger names with t_.

tables:

t_<tablename>_iud or t_<tablename>_oiud for instead-of triggers.

views:

t_v_<viewname>_oiud

Cheers,

Win

This is not a perfect solution, but allows you to test completed stages of the tirgger logic.  It is also more complicated to implement. The trigger must fire or it...

Clay,

I run a 2-line script with dbcc dropcleanbuffers and freeproccache before every benchmark test. Read - ahead count is also the same.

cheers,

Win

You can do the opposite (cache) and keep the table in the database buffer cache with dbcc pintable, but I am not aware of a nocache equivalent.

Win

Steven,

Thanks for the proc, let's hope there will be a command in Yukon

Win

Brian,

Is there some undocumented - hard to find command to disable instead-of triggers on views - apart from dropping the trigger?

Thanks,

Win

Thanks for your comments - this is not the first time someone made a comment about the non-uniqueness of ROWGUID. 

Cheers,

Win

Thanks guys. This is a cut and paste from notepad. Twice burnt....

This database forms the back-end to a commercial data management system.  The front-end is written in C++ and sits...

Note to the coordinator of the thread.  I just wrote a lengthy reponse to Kenneth's comments.  I pressed the preview button and  lost the whole opus.  This is the second...

If what you are saying is correct SQL SVR is not optimised to handle a ROWGUID  better than any other 16 byte datatype.  The integer Identity looks more attractive.  It avoids...

1) I assume your sequence exists

2) I cannot find any declaration of

sysdt date;

nnx  number;

in package, package body or procedure

for

select sysdate into sysdt from dual;   -- SQL Server equivalnet is:   select @sysdt=getdate()...

 You can use temp tables to test your design if up-front re-coding looks like too much work.  Temp tables created inside a procedure are automatically destroyed when they go out of...

Why don't you create the datatype table following the same principle and stick with the information_schema views which are more up-to-date than the old hack of using sys...tables.

Win