Article Preview
Top1. Introduction
Internet of Things (IoT) has been largely adopted as a network paradigm (Liu et al., 2021). Query processing of the huge volume of heterogeneous data generated by a big number of IoT devices is one of the most important challenges in IoT (Widya et al., 2018). In contrast to traditional data management systems, in which data is structured and query languages are standardized, each IoT device has their own data processing technique (Widya et al., 2018). IoT paradigm is a set of devices equipped with identification, detection, processing and networking functions (Diène et al., 2020). Such devices communicate with other Internet devices and services to perform specific tasks. IoT paradigms play a fundamental role in exchange and communication in large-scale networks (Savaglio et al., 2020). Data management in the IoT paradigm is based on solutions providing efficient storage and indexing of structured and unstructured data (Diène et al., 2020).
In other hand, to increase performance and availability in data management systems, data is fully or partially replicated using various strategies through caching and view materialization techniques (Sebaa and Tari, 2019a). Such techniques play an essential role in data replication and efficient query processing. In distributed data management systems, materialized views can be stored in different nodes like base relations. They can improve query performance through query rewriting for frequently issued queries and increase availability in case of network failures (Raipurkar et al., 2021). However, the problem of managing materialized views at particular nodes becomes more difficult in distributed systems. Indeed, in such an environment, views management becomes more challenging, since other issues such as decentralized control of data, resources restriction, symmetric communication, unmanaged update, query patterns, and relationships between views must be considered. Furthermore, resource restrictions, such as CPU, I/O, and network bandwidth must all be considered. (Bellahsene et al., 2010) Moreover, the most difficult task of materialized view management is the maintenance process. Thus, data management applications tend to the materialization of the most requested data portions and their replication across multiple nodes to reduce query run time (Sebaa and Tari, 2019b; Sebaa et al., 2019). Therefore, the probability to find views referencing one or more common sources is great. However, this situation complicates the maintenance task of materialized views and their replicas (Zhou et al., 2007; Sebaa et al., 2019).