I have done the following extra credit work:

--included the min/max statistics
--7c
--7d

All three simulations are available by running Simulation.java



After running all three simulations on an input data file, 
it seems that one line with one teller is very slow. With 
my last data set, I had an average wait of 22, with a high 
of 59.  With the same events, 7c and 7d each had an average 
wait of only 6 and highs of 22 and 21 respectively.  

I would hazard that three windows with one queue/line is the 
most efficient.  Having one line/one window obviously makes 
everyone wait when one person has a long task.  The same 
occurs to a smaller scale in each of the three queues when 
there is a long task -- everyone behind that person in that 
queue has to wait, even though other queues may be moving faster.
(I think we all know that experience!)  

Were this a real simulation, I would want to run data collected 
from the actual bank--do they have a lot of long tasks, or a 
lot of short tasks?  Also, one has to keep in mind the cost of 
extra tellers, and whether it is simpler to make queues for 
different types of tasks, rather than each teller handling 
all inquiries.  People are probably willing to wait longer when
they know their own tasks will also take a long time.

  