The verification tolerance, , is the amount of difference allowed between the LP state and the actual entity state. A large tolerance decreases the number of false messages and rollbacks, thus increasing performance and requiring fewer queries, but allows a larger probability of error between predicted and the actual values will cause rollbacks in each LP at real times of from the start of execution of each LP.
The error throughout the simulated system may be randomized in such a way that errors among LPs cancel. However, if the simulation is composed of many of the same class of LP, the errors may compound rather than cancel each other. The tolerance of a particular LP, , will be reached in time . The verification query period ( ) should be periodic with period less than or equal to in order to maintain accuracy within the tolerance.
The accuracy of any predicted event must be quantified. This could be quantified as the probability of occurrence of a predicted event. The probability of occurrence will be a function of the verification tolerance, the time of last rollback due to verification error, the error between the simulation and actual entity, and the sliding lookahead window.
Every LP will be in exact alignment with its PP as a result of a state verification query. This occurs every time units.