Management of the Real-Time Derived Data in a Distributed Real-Time DBMS Using the Semi-Total Replication Data Sana Hamdi, Malek Ben Salem, Emna Bouazizi, Rafik Bouaziz MIRACL Laboratory, Higher Institute of Computer Science and Multimedia, Sfax University, Tunisia Email: hamdisana@gmail.com, bensalemmalek@gmail.com, emna.bouazizi@gmail.com, Raf.Bouaziz@fsegs.rnu.tn Abstract—In real-time databases, there are data, called sen- sory. They are regularly updated by periodic update transactions. These data are used to calculate other data, called derived data. At each update of a given sensory data, the derived data which depend on it should be updated. Many studies have been conducted to design architectures for Distributed Real-Time DBMS (DRTDBMS) but excluding real-time derived data. In this paper, we study various policies for updating derived data and apply a mixed method for updating derived data based on a management approach to Quality of Service (QoS) using a semi- total replication as a principle of data distribution. Keywords—DRTDBMS, Quality of Service, Derived Data, Real- Time Transactions, Semi-Total Replication. I. I NTRODUCTION Many environments of real-time applications are dis- tributed. Processing the information of such systems requires a consideration of the distribution of data and processing. Centralized architectures of RTDBMS appear insufficient to be used in distributed systems. Therefore, the use of RTDBMS becomes an essential solution. The DRTDBMS include a collection of sites connected together via communications’ networks for transaction pro- cessing. The presence of several sites poses problems that were not present in centralized and RTDBMS and DRTDBMS performance depends on the distribution of the workload between sites. In a distributed real-time application based on the use of DRTDBMS, the DRTDBMS are qualified by the total distribution of transactions and data replication. There are two types of real-time data [5]: • Base data (sensory): it is usually data from sensors that report the state of the environment • Derived data: they are derived from one or more base data and used to deduce information about the environment. Transactions submitted by users arrive at varying frequen- cies. As the frequency increases dramatically, the balance of DRTDBMS is jeopardized. During these periods of overload, DRTDBMS will potentially run out of resources and real- time transactions will then miss their deadline in greater numbers. To manage these phases of system instability (phase overload or phase in use), work-based QoS approach tries to make more robust DRTDBMS but regardless of derived data. This work is based on techniques of scheduling with a feedback control called DFCSA (Distributed Feedback Control Scheduling Architecture). Our contribution consists in adding the notion of real- time derived data in DFCSA. In this article, we begin with a presentation of existing approaches on which is our based work. Then, in section 3, we present the model we consider. Section 4 shows the details of the simulation settings and the evaluation results. We conclude this article by a discussion of this work and a presentation of our future works. II. RELATED WORK There is a growing need for real-time data services in distributed environments. The issues involved in providing pre- dictable real-time data services in centralized database systems have been researched and the results are promising. However, we are not aware of research results for providing data services with Quality-of-Service(QoS) guarantees in distributed real- time database environments [10]. A. Distributed Real-Time Database Model 1) Data Model: Data objects are classified into either real- time or non real-time data. A non real-time data is a classical data found in most of databases, whereas a real-time data has a validity interval. These data may change continuously to reflect the real world state (for example, the current temperature value). Each real-time data has a timestamp indicating the last observation of the real world state. Data replication is an interesting alternative to a DRT- DBMS as the availability of data on several sites could help database systems meet the stringent temporal requirements of real-time applications [6]. Data replication greatly improves the system performance when the majority of operations on data replicas are read operations. It helps also avoid the data 978-1-4799-0792-2/13/$31.00 ©2013 IEEE