D ESCRIPTION : Performance tuning is difficult without a solid baseline or
when multiple configuration parameters are changed at once between
measurements. Attempting performance tuning in these conditions can
result in similar or identical tests run several times, giving the appearance
M OST FREQUENT SCALE : Application.
R EFACTORED SOLUTION NAME : Good Performance Methodology.
R EFACTORED SOLUTION TYPE : Process.
R EFACTORED SOLUTION DESCRIPTION : Good testing methodology and a
good testing environment are the primary keys. Baseline measurements
are mandatory. All tests should start from a common configuration and
change a single parameter at a time.
R OOT CAUSES : Haste, ignorance.
A NECDOTAL EVIDENCE : “What did we change for that last test?” “Didn't
we just do that?” “We're right back where we started.”
S YMPTOMS , CONSEQUENCES : Inefficient performance testing and tuning,
longer-than-expected performance tuning cycles, and unclear results of