Determine performance issue

I read from a book and part of the comment, it mention this:
======================== After further analysis, we determined that more than
80 percent of the total resources being consumed-processor, memory,
and
input/output (I/O)-was involved in this conversion process.
========================
The conversion process in the comment is the conversion of datetime
format to another format.
QUESTION:
1. How can we determine the percentage of time was involve in
conversion process?
2. How can we check for it? Using profiler?
I'm interested to the method of checking performance problem, but have
no idea where to start with and what to check.
Any guideline?
Thanks.
wodoy.peterOne reliable method is to conduct controled tests to compare the performance
between using the conversion and not using the conversion. This may not be
possible in some scenario, and may be too expensive in some other scenarios.
Without such a comparison, I'm not sure how one can determine with certainty
that X% of resources is being consumed by a conversion step by looking at
perfmon counters or SQL Trace results.
Linchi
"wodoy.peter" wrote:
> I read from a book and part of the comment, it mention this:
> ========================> After further analysis, we determined that more than
> 80 percent of the total resources being consumed-processor, memory,
> and
> input/output (I/O)-was involved in this conversion process.
> ========================> The conversion process in the comment is the conversion of datetime
> format to another format.
> QUESTION:
> 1. How can we determine the percentage of time was involve in
> conversion process?
> 2. How can we check for it? Using profiler?
> I'm interested to the method of checking performance problem, but have
> no idea where to start with and what to check.
> Any guideline?
> Thanks.
>
> wodoy.peter
>|||On Fri, 2 Feb 2007 14:12:01 -0800, Linchi Shea
<LinchiShea@.discussions.microsoft.com> wrote:
>One reliable method is to conduct controled tests to compare the performance
>between using the conversion and not using the conversion. This may not be
>possible in some scenario, and may be too expensive in some other scenarios.
Excellent point, just comment out the line and use the wall clock!
When possible ...
>Without such a comparison, I'm not sure how one can determine with certainty
>that X% of resources is being consumed by a conversion step by looking at
>perfmon counters or SQL Trace results.
Profiler can give you some pretty fine-grained results, if you turn
everything on. It also impacts performance, especially at those
levels.
J.