Optimal Policies for Database Error Clearing
김기수 · 1993
This paper models the accumulation of errors in a database using a stochastic clearing system with a generalized liner cost structure. Using this model, we show that a control-limit policy for removing errors is preferable to a fixed time interval policy. Since the number of errors in the database generally cannot be observed continuously, we develop and evaluate an opportunistic clearing policy which assumes the number of errors can be observed once per clearing cycle. An approximation is developed for the optimal time at which to observe the number of errors. Numerical studies of this policy show it to be superior to the fixed time interval policy.