Big data analytics for network and service management

Abstract

In recent years, we observe a growing trend in the area of big data analytics for network and service management. Approaches such as statistical analysis, data mining, and machine learning have become promising to harness the immense stream of operational data and to improve operations and management of information technology systems and networks. Huge amounts of data from data archives, data centers, cloud systems, Internet of Things, and the Internet are collected, shared, and analyzed for the management of the networks and services. The features of big data, namely, volume, variety, velocity, and veracity, bring new challenges to network and service management. To manage the configuration, performance, resilience, availability, and security of the networks and services, traditional measures such as log/event analysis, intrusion detection/prevention, and monitoring and deployment have taken a new dimension. New techniques and mechanisms from machine learning, data mining, and data visualization are explored for designing, developing, and operating services and networks in the big data era. In summary, there are a lot of research challenges in this emerging field of data analytics. The purpose of this Special Issue is to explore and highlight the promising capabilities of data analytics in managing the huge streams of operational data on the networks and services. Thirteen papers were submitted for this Special Issue. Four of them are the extended versions of the research presented at the 2016 IEEE/IFIP International Workshop on Analytics for Network and Service Management. After extensive reviews and discussions, six papers were finally selected for publication in this Special Issue. The authors of these papers were given the time to update their papers based on the review comments and suggestions provided. The selected papers address topics that play a central role in using big data analytics for network and service management and presenting novel theoretical and/or experimentation results. The first paper, “Data Transformation as a Means towards Dynamic Data Storage and Polyglot Persistence,” by Vanhove et al is on dynamic storage solutions for big data centers. The authors propose a transformation approach through a canonical model based on the Lambda architecture. The proposed solution is evaluated through a network monitoring platform considered as a use case scenario. The authors also discuss the support for SQL, CQL, MongoDB, and NoSQL document data stores. The second paper, “Dynamic Resource Allocation for Big Data Streams based on Data Characteristics (5Vs),” by Kaur et al proposes a learning based system for dynamic resource allocation using the characteristics extracted from the big data streams. Self Organizing Map, a neural network-based clustering technique, is employed and experimentally evaluated. Comparisons with other systems such as Storm, Flume, and S4 demonstrate the benefits of the proposed system. The third paper, “Learning Ensemble Strategy for Static and Dynamic Localization in Wireless Sensor Networks,” by Ahmadi et al explores the usage of ensemble learning for the purpose of indoor localization in wireless sensor networks. The proposed approach has been experimentally evaluated using real measurements and compared with other learning-based localization algorithms currently available in the literature. The fourth paper, “ALACA: A Platform for Dynamic Alarm Collection and Alert Notification in Network Management Systems,” by Solmaz et al presents the design and development of a scalable streaming alarm management system that includes complex event processing. The proposed system is designed to work on operation support systems that are used by the mobile network operators. Based on evaluations and experimental results, insights are provided on the performance of real-time alarm data analytics techniques and tools. The fifth paper, “Security in IoT Network based on Stochastic Game Net Model,” by Kaur et al is on the security issues and counter measures of Internet of Things. The authors propose a stochastic game net–based model for the security of Internet of Things where they combine the advantages of stochastic Petri Nets with game theory. The proposed approach is evaluated, and experimental results are presented to characterize their effectiveness and efficiency. Finally, the last paper, “Botnet Behaviour Analysis: How would a Data Analytics-based System with Minimum a priori Information Perform,” by Haddadi et al analyzes different machine learning techniques for network traffic–based botnet behavior identification. Botnets being one of the biggest threats in cyber security, several works in the literature proposed data analytics–based solutions. The authors evaluated and discussed five most popular techniques on 24 publicly available data sets in terms of their performance, cost, and a priori information requirements. We expect future work in the area of big data analytics for network and service management to further investigate the topics addressed by the selected papers. The guest editors of this Special Issue wish to acknowledge the excellent work that has been performed by the authors, who submitted papers, and by the reviewers, who have spent a considerable amount of their time providing high-quality reviews. We would also like to extend our thanks to the editorial board of the International Journal of Network Management, in particular to Editor-in-Chief James Won-Ki Hong and Associate Editor-in-Chief Filip De Turck, for the great opportunity and their support in editing this Special Issue. Last but not the least, we thank the Editorial Team of Wiley, for the support offered to the authors.