No open-access location is known for this record yet.
No open-access location is known for this record yet.
There has been a tremendous increase in both the capacity and complexity of data in the last few years, and for this reason, big data technology has been one of the most conspicuously emerging research areas. In big data technology, communication systems are implemented in such ways that data are collected from heterogeneous services and integrated seamlessly. To make big data technology more user-centric, further exploration is called for in the areas of data utilization and communication system in heterogeneous network environments 1-4. Therefore, many scientists have been actively engaged in researching in the following fields in connection with big data technology: networking, data mining, information security, and privacy protection. This special issue aims to focus on the topics of communication system in enhanced big data technology and application 1-4. Areas of interest for the special issue ‘Advanced Communication Systems for Enhanced Big Data Technology and Applications’ include the following topics: big data processing and analytics in communication systems; modeling, semantics, and protocols for big data; agent technology and context-aware system; big data mining for network optimization; social networks influence on communication systems; real-time and streaming data processing and analytics; network protocol for ubiquitous data gathering; data management for mobile and pervasive computing and networking; architecture and optimization of advanced big data technology; cloud/grid/stream computing for big data; applications and security for big data; testbeds and case studies for big data technologies and applications. We have received many manuscripts. Only six manuscripts with high quality were finally selected for this special issue. Each manuscript selected was blindly reviewed by at least three reviewers consisting of guest editors and external reviewers. We present a brief overview of each manuscript in the following. The first paper entitled ‘MR-tree: an efficient index for MapReduce’ by Chunsheng Li et al. proposes a novel method, called MR-tree, to index multiple dimensional data in MapReduce framework 5. The generation of an MR-tree contains two phases. In the first phase, it creates the leaf-level index layer. In the second phase, it uses an iterative manner to build internal-level layers. The main contribution of this paper was MR-tree, a tree-like structure to index multiple dimensional data. The construction of an MR-tree contains two phases, namely, leaf-level building phase and internal-level building phase. All entries in leaf layer are constructed in the first phase, whereas the rest of the entries are constructed in the second phase. Experimental results indicate that the proposed method can run efficiently. The second paper entitled ‘Two-phase grouping-based resource management for big data processing in mobile cloud computing’ by JiSu Park et al. proposes a grouping technique based on the utilization and movement rates 6. In this scheme, mobile devices are separated into groups by cutoff points based on entropy values. Authors also proposed a two-phase grouping method in order to reduce the overhead of group management. Authors proposed the two-phase group classification in a variety of distributions of mobile device. The existing studies that provide arbitrary cutoff points are not appropriate for mobile cloud environments, which have the volatility issue. The proposed method provides a two-phase grouping by integrating groups from entropy-based grouping with reflecting the similarity between the groups. The experimental result shows that the algorithm generates two-phase groups successfully even though the distribution of mobile devices changes. In the first experiment, we measured the time for the entropy calculated using various number of cutoff points per factor. We also compared the average fault rates of various numbers of cutoff points per factor. The result shows that the optimal number of cutoff points depends on the distribution of underlying mobile devices. Therefore, the number of cutoff points should be chosen considering the characteristics of underlying mobile devices. In the second experiment, we measured the execution time of classified groups with one-phase grouping and two-phase grouping. We also compared two grouping techniques with different group selection algorithms. The result shows that the two-phase grouping technique outperforms the one-phase grouping technique regardless of the group selection algorithm. Another paper in this special issue, entitled ‘Unstructured deadlock detection technique with scalability and complexity–efficiency in clouds’ by JongBeom Lim et al. presents an unstructured deadlock detection algorithm using a gossip protocol in cloud computing environments, where constituting nodes may join and leave at any time. Because of the inherit properties of a gossip protocol, authors argue that our proposed deadlock detection algorithm is scalable, fault-tolerant, and efficient, retaining safety and liveness properties 7. Authors proposed an unstructured deadlock detection algorithm using gossip communication to cope with scalability, fault tolerance, and complexity–efficiency issues. A cloud environment, where the behavior of the constituent nodes is active and dynamic (i.e., nodes join and leave at any time), is an example of an environment in which our algorithm will be applied. Unlike the algorithms proposed in previous studies, our algorithm shows that messages generated are diffused among nodes almost evenly without a bottleneck and that the requisite number of cycles increases linearly as the number of nodes grows exponentially. The message complexity of the algorithm is O(n), where n is the number of nodes, satisfying the safety and the liveness properties. In addition, our deadlock detection algorithm could be embedded seamlessly into other existing gossip-based algorithms. In other words, if a gossip-based algorithm is implemented to provide a failure detection service, the deadlock detection algorithm proposed in our work can be embedded in the existing gossip-based algorithm. The fourth paper entitled ‘IoT-SVKSearch: a real-time multimodal search engine mechanism for the internet of things’ by Zhiming Ding et al proposes a real-time multimodal search engine framework, IoT-SVKSearch, for retrieving massive and heterogeneous Internet of Things (IoT) sampling data 8. Through the distributed global indices built on the extracted full-text keywords, the spatial–temporal attributes, and the sampling values, IoT-SVKSearch can support multi-modal search conditions including keyword-based, spatial–temporal, and value-based constraints. Next paper entitled ‘An analysis of performance factors on Esper-based stream big data processing in a virtualized environment’ Mino Ku et al. analyzes four performance-influencing factors on a virtualized event processing system: the number of query statements, the garbage collection intervals, the quantity of virtual machine resources, and the virtual CPU assignment types 9. Through the result of the first experiment, this paper suggested that the parallelization of event processing processes as a function of the number of query statement is necessary. Through the parallelization, we can make an event processing system to satisfy a processing time requirement and efficiently manage the memory space of event processing systems. In the second experiment, it could check that appropriate intervals of the Garbage Collection (GC) executions helped to manage memory efficiently, with minimized performance loss. Also, periodical GC executions had the effect of making a memory usage boundary under event processing. In the third experimental result, we could check that the lack of memory caused significant performance degradation rather than the lack of vCPUs, due to CPU usage characteristics of an event processing engine, that is, Esper. In particular, supplying additional vCPUs to a virtual machine that has insufficient memory spaces aggravated the event processing performance. Finally, in the fourth experimental result, vCPU assignment types directly affected the event processing performance; throughputs dropped by 20–33% as a function of vCPU assignment types. The last paper entitled ‘Task scheduling scheme based on resource clustering in desktop grids’ by Joon-Min Gil et al. proposes a task scheduling scheme based on resource clustering that can selectively allocate tasks to those resources that are most suitable for the current situation of a desktop grid environment 10. As a classifier of resources, the k-means clustering algorithm is introduced to classify resources according to their own task execution availability and result-return probability. This paper presented a task scheduling scheme based on resource clustering, which can selectively allocate tasks to those resources that are suitable for the current environment of a desktop grid. The proposed scheme incorporates task execution availability and result-return probability to capture the execution behavior of resources appropriate to the current situation. It also uses the k-means clustering algorithm to classify resources into resource groups. Our scheduling scheme can thereby achieve efficient task scheduling by aggressively taking into account the execution behavior of each resource group in the process of task allocation. Our special thanks go to Prof. Mohammad S. Obaidat and all editorial staffs for their valuable supports throughout the preparation and publication of this special issue. We would like to thank all authors for their contributions to this special issue. We also extend our thanks to the following external reviewers for their excellent help in reviewing the manuscripts.
Loading…
Stored only when both papers are in the corpus — 6 citations exist upstream.
Papers that cite the same references — the relational fallback while embeddings are still backfilling.
Imported from OpenAlex (CC0) — nothing here is generated.