Temporal Consistency Maintenance for Real-Time Update Transactions

A real-time data object, e.g., the speed or position of a vehicle, or the temperature in an engine, is temporarily consistent if its value reflects the current status of the corresponding entity in the environment. This is usually achieved by associating the value with a temporal validity interval [15,16,14,10,19,20,11,3,5,21]. For example, if position or velocity changes that can occur in 10 seconds do not affect any decisions made about the navigation of a vehicle, then the temporal consistency interval associated with these data elements is 10 seconds. One important design goal of real-time and embedded database systems is to always keep the real-time data temporally consistent. Otherwise, the systems cannot detect and respond to environmental changes in a timely fashion. Thus, sensor transactions that sample the latest status of the entities need to periodically refresh the values of real-time data objects before their old values expire. Given temporal consistency requirements for a set of sensor transactions, the problem of designing such sensor transactions comprises two issues if transactions are scheduled periodically [20]: (1) the determination of the updating period and deadline for each transaction from their temporal consistency requirements; and (2) the scheduling of periodic sensor transactions. Minimizing the update workload for maintaining temporal consistency is an important design issue because (1) it allows a system to accomodate more sensor transactions; (2) it allows a real-time embedded system to consume less energy; and (3) it leaves more processor capacity to other workload (e.g., transactions triggered due to detected environmental changes). To monitor the states of objects faithfully, a real-time object must be refreshed by a sensor transaction before it becomes invalid. i.e., before its temporal validity interval expires. The actual length of the temporal validity interval of a real-time object is application dependent. Sensor transactions have periodic jobs, which are generated by intelligent sensors that sample the value of real-time objects. One important design goal of Real-Time Data Bases (RTDBs) is to guarantee that real-time data remain fresh, i.e., they are always valid. Temporal consistency (a.k.a. absolute consistency [15]) is proposed to determine the temporal validity of a real-time data object based on its update time by assuming that the highest rate of change in a real-time data object is known.