The data contained herein is the data that is graphed in our paper.
For more data (using older algorithms, different combinations of
parameters, or whatever), talk to Kent Hunter.  He will be happy to
provide you with a CD (or CDs) containing all the data that were
collected and the scripts used to collect them and sift through them.

This directory contains essentially four different kinds of
subdirectories (each containing a separate file for a number of
different settings of the "percentage of blocks initially free"
parameter):

syssim-simple-out-* contains output using our original, simple
algorithm.

syssim-simple-out-0.5r-* contains output using our original, simple
algorithm on input files in which each block has a 50% chance of being
the immediate successor of the one before it.

syssim-out-* contains output using our more complex (but, it seems,
less efficient) cylinder group algorithm.

syssim-goodneg-angle-out-* contains output using a new version of the
simple algorithm that looks at block positions rather than block
numbers to determine rotationally optimal block placement.  (This is a
slight improvement over an earlier algorithm that handled negative
angular displacements incorrectly--hence the "goodneg".)

The suffixes Xs.X on each subdirectory indicate that the files in that
directory were produced from input containing X blocks to be written.
(Originally, the number before the 's' was size of groups that were
sorted in ascending order, and the second number was the total number
of requests, but for the data here, these numbers are always the
same.)
