I wrote an article about the problem in 2006. Besides the solution with table function, look at the third solution: http://www.sql-server-performance.com/articles/dba/list_random_values_p1.aspx

I still use SQL Server 2000 and for that version execution plan is created when the...

Thanks again for stopping by and for the constructive comments. I'd also love to see your article (as I'm sure the others on this discussion would) on the subject....

I agree that it depends and there are cases for doing implementation in SQL.

However, to clarify my position: One should consider sql implementation of running totals only if...

Hello Jeff & Co. Really great article and awesome discussion. I wrote an article about the topic five years ago, but my investigation lacked depth of the Jeff's article and...

See: http://www.sql-server-performance.com/forum/topic.asp?TOPIC_ID=14259

The best solution is to return raw data set and to calculate cumulative values on the client. For more details read: http://www.sql-server-performance.com/mm_cursor_friendly_problem.asp

Curosor solution is most probably faster then one mentioned above.

Another approach is to build table variable (or temp table) and with tabl21IDs column initally null. Then you can update...

Jeff, I agree with your discussion about row-by-row processing, however I disagree about temp tables vs. table variables. My points are:

1. I don't need statistics on temp tables, because I...

Again article suggesting replacing a cursor with one row at time processing loop. That's not more efficient solution. It would be nice if author would test the solution before claiming...

I tested the loop solution and it hardly offered any performance improvement over cursor. I would like to have test specification that showed 75% improvement.

I already posted the comment that was here once I checked, but dissapeared in the meantime.

Anyway, I like the article, it is pretty good beginners guide to index design.

However, I...

Using union instead of or is good idea. On top of that index on modifiedByUserID would help peformance.

Clustered index on createdUserID causes a lot of page splits during inserts. Consider...

I think you wanted to say that if clustered index is not on identity column that can cause a lot of page splitting. Anyway, I agree we need much more...