We comment here on the analysis required for the data samples discussed above. Potentially the largest sample will be the 1200 hour run for the positive strangelet search. If the trigger is sufficiently loose so as to saturate the DA system (a situation we do not expect), then the run could produce:
The analysis programs used to study the Monte Carlo data discussed above do not have to
deal with the full complexity of the detector, since our simulation is still evolving.
However we have made a very realistic simulation of the scintillation counter hodoscopes
and, as discussed in the section on the simulations, the hodoscopes form the basis for the
pattern recognition. Currently, the full tracking analysis for events with a HIJET
central interaction, tracks from interactions in the shielding and a strangelet takes
0.215 seconds per event on a VAXstation 3100 M76 (9 VUP's). To do only the hodoscope
pattern recognition takes 0.105 seconds per event on the same machine. Since the
hodoscope tracks give the velocity measurement and a mass measurement good to about
10% ( 
) we will be able to reject most events with only the hodoscope
tracking. This means that the main data reduction for the 1200 hour run will not take more
than 13.7 VAXstation 3100 M76 years. On the CPU power currently available in the collaboration
this would take about one calendar year. We fully expect that the current rate of
evolution of CPU power will allow us to have a least a factor of four more CPU power
available by the time this experiment has data. Further, the first runs will
most likely not be the long high sensitivity runs. By the time that analysis is
required it is reasonable to expect that it should not take more than a few months
calendar time on the CPU power available in the collaboration. Thus we do not see
a need for major expenditures for specialized processing for this experiment.
We also note that the currently available storage medium of choice (8mm tapes) will hold
about 
events each. At 0.1 sec per event this means a VAXstation 3100 M76 will
take about 14 hours to process one tape. At full transfer rate the 8mm tape drives can
read an entire tape in 2 to 2.5 hours. Thus even with processors 5-7 times faster, the
analysis will still be CPU bound so new technology is not needed to match the data
storage medium to the processing speed. Of course we will take advantage of any
appropriate improvements in storage technology that become available.