Stochastic Analysis of Mechanizing Transaction Data Bases

In this paper, we analyze the generic problem of implementing a system that requires manual effort for converting paper records in a manual file system (MFS) to mechanized computer records in a data base management system (DBMS). We can achieve conversion to a mechanized environment in a number of ways: by going-forward on an activity basis, by crash-conversion utilizing an augmented work force, or by some mix of the two dictated by available resources such as money, people, space, and time objectives. A paper record is mechanized using the going-forward approach if the record is transferred from the MFS to the DBMS by regular record 91 maintenance employees when a change or deletion activity is performed upon that record. If a randomly chosen record is transferred by a separate temporary work force, the record is said to be transferred using the crash approach. The paper record could contain all of the information that is intended to be mechanized or an enhancement to an already existing mechanized record. The arrival times of activities (change or deletion orders) for a particular record are assumed to form a Poisson process. We further assume that the distributions of the activity arrival times for every record are independent and identical. Hence, the sequence of times at which these change and deletion activities occur forms a nonstationary Poisson process. In addition, the transfer of records is carried out by a special task force with constant total mean rate, the amount of work required for a record to be transferred being exponentially distributed.