Article Preview
Top1. Introduction
Big data (e.g., (Zikopoulos & Eaton, 2011; Manyika et al., 2011; Cuzzocrea, 2013b; Cuzzocrea et al., 2011; Li et al., 2015)) are now of critical interest for both academic and industrial research communities. The well-known characteristics of big data are representing a big challenge that is being faced-off by means of several research perspectvies and angles (e.g., (Bessis & Dobre, 2014; Huang et al., 2013; Reina et al., 2015)).
Among several problems, the issue of effectivly and efficiently supporting big data management plays a critical role (e.g., (Yang et al., 2014)). Indeed, classical approaches cannot deal with the tremedous amounts of data produced by actual data-intensive platforms, such as social networks (e.g., (Peng et al., 2018)), smart cities (e.g., (Massobrio et al., 2018)), experimental platforms (e.g., (Feng et al., 2018)) and so forth.
In our opinion, one of the most problematic characteristic of big data is represented by their heterogeneity, as typical real-life big data repositories include data sources of several types, such as geo-spatial data, 3D data, audio and video data, unstructured data, graph data, and so forth. This is, indeed, a critical challenge for next-generation big data applications and systems, for instance those based on smart cities.
Deadling with big data means dealing with powerful architectures that aim at creating the so-called “big data ecosystem” (e.g., (Chen et al., 2018)), a reference architecture that supports all the phases that are necessary to effectively and efficiently support big data representation, storage, management and mining. Within these proposals, distributed environments are clearly the most representative setting where the effectiveness and efficiency goals over big data can be achieved successfully. In this respect, (Kennedy et al., 2019; Chon et al., 2018) are some noticeable experiences.
In order to cope with such issues, several approaches have been proposed recently. For instance, a class of approaches among the most popular ones predicates to adopt well-consolidated data compression methods (e.g., (Sayood, 2017)) for reducing the effort of accessing, processing and managing big data.
This line of research has been traditionally adopted by the database management research community, and now it is of great interest for the hereditary research community that focuses the attention on big data management.
On the other hand, it is a clear evidence that, since the last few years, the dramatic increase of data generation rate has been issuing new research challenges for data management, mainly falling in the emerging context of big data, like also highlighted by several studies (e.g., (Bellatreche et al., 2010; Manyika et al., 2011; Zikopoulos & Eaton, 2011; Cuzzocrea et al., 2013a; Cuzzocrea et al., 2013b)). There are several scenarios (such as control systems for climate disaster prevention, traffic network monitoring systems, etc) where huge amounts of data produced by sensor networks are collected and queried to support both punctual and trend analysis. In order to make sensor data analysis feasible, new data representation techniques and querying algorithms are required. In fact, traditional approaches, mainly coming from the Database research area, cannot efficiently manage sensor network data, as the stream of sensor network readings is theoretically unbounded. To support qualitative and statistical analysis, aggregate queries are issued on sensor readings. One of the most relevant research issues in this context is to make answering aggregate queries as much effective and efficient as possible. For instance, environmental sensor networks supporting climate disaster prevention could exploit fast answers of queries on environmental parameters in order to provide a timely reaction to the world.