RE: How To Compare Two Processes?

[Steven_S_Prevette@RL.gov]

It sounds like there are two variables to be measured in your problem.  

The first is the failure rate of each system.  A single failure is defined
as the random test sends a signal to the system, and receives no response
(or "times out", I assume it does not wait forever).  The failure rate would
be the number of failures divided by the total tests sent.  Each system
should have this variable plotted on a p-chart.  The rate could be plotted
by day, assuming there are at least say 10 tests per day per system.

The second variable appears to be response time, given that a response was
received.  No response would simply be "no data", and the datum point for
that time would not be plotted.  You could leave a gap in the graph if
desired.  The control chart would be an xbar-R set, or an x-chart.

It may be a good idea to test both systems simultaneously, rather than
running independent random number generators on each, in case there is some
other factor effecting system performance.  You could stack the two charts
above each other.  This is where leaving gaps in the xBar-R chart for the
second variable would pay off, so that the timelines will remain in synch.
Also, the control charts will tell you if each system is statistically in
control, and only if the two systems are in control (or at least the
non-modified system is in control) should there be a comparison made.  I
would compare the system by comparing the two averages on the control
charts, and determining the standard deviation of the AVERAGE line (not
individuals) for the comparison.  

Steve Prevette
ESH Planning and Performance
Fluor Hanford, A Fluor Global Services Company
ASQ Certified Quality Engineer
steven_s_prevette@rl.gov
509-373-9371
========================================================================