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

On 28 May 2004 18:14:50 -0700, Neo wrote:

Hi Neo,

Even though completely irrelevant to the challenge, I'll take the time to addrress your questions.

>> I changed my model ...
>> the average execution time is now down to 11.0 ms.
>
>XDb1 v4.5.4 takes 0.973 msec to generate the report in RAM only (does
>not include writing the report to hard disk). With SQL Server, can one
>generate the report to RAM only and measure the time?

No. SQL Server is designed for commercial use. Most companies would never use a product that might lose data if the power fails. SQL Server is designed to make sure that after a power outage, all transaction that were committed at the time of the outage will still be complete after restarting (even if the completion was only a fraction of a millisecond before the outage). The concept of RAM tables is not compatible with the concept of no data loss at power outage.

> When running the
>report via Query Analyzer, does the reported time include writing
>table NCancestor to hard disk?

It does include the time to write the log to hard disk. The log contains all the changes made to the ancestors and the NCancestor tables. The data pages of those tables themselves may or may not be written in that time; SQL Server has a background process that will periodically flush "dirty pages" from cache to disk.

>Apps: XDb1 (v4.5.4) and Notepad.

Ouch - now you are even increasing the version number in between two messages. Yikes.

Best, Hugo

-- 

(Remove _NO_ and _SPAM_ to get my e-mail address)