Dynamic on-demand updating of data in real-time database systems

The amount of data handled by real-time and embedded applications is increasing. Also, applications normally have constraints with respect to freshness and timeliness of the data they use, i.e., results must be produced within a dead-line using accurate data. This calls for data-centric ap-proaches when designing embedded systems, where data and its meta-information (temporal correctness require-ments etc) are stored centrally. The focus of this paper is on maintaining data freshness in soft real-time embedded systems and the target application is vehicular systems. The contributions of this paper are three-fold. We (i) define a specific notion of data freshness by adopting data similar-ity in the value-domain of data items using data validity bounds that express required accuracy of data, (ii) present a scheme for managing updates in response to changes in the data items; and (iii) present a new on-demand schedul-ing algorithm, On-Demand Depth-First Traversal denoted ODDFT, for enforcing data freshness by scheduling and ex-ecuting update transactions. Performance experiments show that, by using our updating scheme and introduced notion of data freshness in the value-domain, computational work im-posed by updates is reduced for both the new ODDFT and well-established on-demand algorithms. Moreover, ODDFT improves the consistency of produced results compared to well-established algorithms. Keywords Real-time database, triggered updates, derived data items, resource management, vehicular systems 1