Re: Nearest Common Ancestor Report (XDb1's $1000 Challenge)

> Small Report Generation Summary (provided by Hugo)
> Platform: 1.3 Ghz PC, Non-SCSI HDs, SQL Server 2000, Windows 2000
> ---------------------------------------------------------------------
> Solution Time(ms) Notes
> ---------- -------- -----------------------------------
> RM's #1 14.3 Db not normalized to atomic symbols
> Avg of 10 consequtive runs.
>
> RM's #2 11.0 Db not normalized to atomic symbols
> Avg of 10 consequtive runs.
>
> XDb1 4.4.7 16 Db normalized to atomic symbols
> Avg of 10 separate consequtive runs.
> Unoptimized, debug version of exe

The challenge is to generate the report from normalized and NULL-less data (I still need to convince you that 'o' in 'john' and 'god' is data). With XDb1, the report is created from normalized and NULL-less data every time, without relying on any intermediate/cached results or statistics from prior runs. Aside from disk I/O, task switching, etc, the first run is as fast as subsequent runs (see consistency of reports generated in RAM only in earlier post). With XDb1, every report generation is "virgin".

However with SQL Server, the average of 10 runs may not be providing the time to generate a "virgin" report. There is a possibility that the last 9 runs are faster because they are relying on intermediate/cached data or statistics from the first run. Can you ensure the time measured is for generating a "virgin" report by using steps similar to below?

1. Restart PC.
2. Create new db.
3. Run script to add data.
4. Measure time for the first run.