The network ancestor attempts to select an attractive node to
send its daughter to, by looking for the highest ratio of
Speed/NumCells. It is hoped that this algorithm will take
advantage of some of the heterogeneity in computing resources
on the net. Figure 6 shows an example of how Speed (CPU speed in
instructions per second) varies over a twenty-four hour period
on one machine in Santa Fe.
Note that the Speed drops to zero
when the user is active, causing the Tierra process to sleep,
which was the predominant condition after the first ten hours.
If genomes migrate to a sleeping Tierra process, their
packets will be lost in the net, and they will die. This should provide a
strong selective force for avoiding migration to sleeping nodes.
While Figure 6 shows the heterogeneity of Speed for one node,
Figure 7 shows the distribution of Speed over a 48 hour period
on a network of 148 machines.
In addition to the high percentage
of sleeping nodes (37%), there is also a wide distribution of
(non-zero) CPU speeds. Organisms on fast nodes can reproduce faster
than organisms on slow nodes, thus there should be selection to
discriminate and migrate to fast nodes.
At the end of October 1997, some bugs were found and fixed in the C code implementing the sensory system. These bugs effectively disabled the sensory system. For example, one bug caused all organisms to only obtain data on several machines, always the same several machines, regardless of how many machines there were on the net. They were blind to about 90% of the net. Although the effects of the bugs were unintended at the time, results will be presented from runs with and without the bug, because the comparison provides an interesting test of the contribution of the sensory system to the fitness of the organisms.