🚀 go-pugleaf

I am testing an application which does batch processing of records in a
database. I run the test against 75 records from the DB; I have two ways of
connecting to the DB that I am trying to compare.  My test fixture is a
VB.Net console application which invokes methods from my compiled dll's.

When I run the two versions (once using each of the two connection methods)
under OptimizeIt's Profiler, I get very suspicious numbers; a 18:1
performance difference.  At first I thought the difference was due to one
method being many times more efficient than the other, but I was suspicious
and ran both outside of the Profiler. If I time the two executions  manually
(on my watch) they both complete in about three minutes--give or take a
second or two.

As far as the Profiler numbers are concerned, running the first version
reports a total processing time of 177,000 ms which is close to the three
minutes I get manually. But the second version reports 10,000 ms --well
below any reasonable measurement I make outside of the Profiler.  Multiple
repetitions of the tests yield similar results, I even created a "Windows
form" test fixture to see if the console application was causing a problem
(the Profiler never launches the application's console; which suggests a
problem and is a bit of a pain, but I'll post that separately)--but the
great disparity between the two execution times remained.

Any idea what is going on here? I hesitate to tweak an application based
upon test results that I find suspicious and are not indicative of execution
times outside of the Profiler (I would expect the ratio of execution times
to remain constant inside and outside of the Profiler; now I am getting an
almost 1:1 execution time of version1:version2 outside of the Profiler and
an 18:1 difference when running inside the Profiler).

FWIW here is the connection code I am changing between the two versions of
the application.  This is buried deep within the execution tree and is not
part of the test fixture console application; I include it simply to dispel
any ideas that the change itself is dramatically affecting the Profiler
(which I would hope this wouldn't):

Version1:
Dim oDs As DataSet
oDs = _oraHelper.ExecuteDataset(myConnectVals.MY_APPNAME, _
                                CommandType.StoredProcedure, _
                                inSP.PKG_SP_NAME, _
                                inSPparms)
'( _oraHelper being an instance of an object from an Oracle supplied
library)

'version2:
'Dim oDs As DataSet
oDs = ExecuteDataset(myConnectVals.MY_APPNAME, _
                     CommandType.StoredProcedure, _
                     inSP.PKG_SP_NAME, _
                     inSPparms)