Guest Editorial: Advances in Parallel and Distributed Computing

Parallel and distributed computing has played a central role in various applications for decades. Recently, with the advances of information and communication technologies, new parallel and distributed systems and applications are emerging at an accelerated pace, promoting this research to a higher level of interest.

In a parallel and distributed system, a computation task is distributed to different servers of the system and processed in parallel at the same time. The development on operating system and underlying software enable the exploding development of parallel and distributed systems. The technique has been used in many large-scale systems, such as datacenter networks, cloud systems, grid system and so on. Several computers can be connected to form a high-performance system by using the technique, which thus own the capability to solve complicated problems. In the chip level, multiple processors can be integrated on the same chip to construct a multi-core computer and work co-operatively for on-site computation tasks. The emergence of these systems and various applications urge the development of new methodologies and techniques for parallel and distributed computing, from processor interconnection and algorithm design to resource allocation and program execution, which will make most effective use of the systems for respective applications.

The special section aims to show some recent advances in parallel and distributed computing, inspired by the extensive scientific discussion carried out at the Sixth International Symposium on Parallel Architectures, Algorithms and Programming (PAAP 2014) held in Beijing, 11-13 July, Beijing. Selected authors were invited to report their latest contributions in parallel and distributed computing in the aspects covering network management, privacy and security, image processing, and mass data computation.

In paper “Distributed locating algorithm MDS-MAP (LF) based on low-frequency signal”, C. Y. Zhou et al. apply the distributed computing method in localization of sensor networks. This paper proposed a multidimensional scale distributed positioning computing methods, which uses low frequency signal to measure the inter-sensor distance rather than shortest path algorithms. The proposed MDS-MAP (LF) algorithm leverages the propagation feature of low frequency signal to acquire a more precisely two-hop distance.

The following paper “An Accuracy Assured Privacy-Preserving Recommender System for Internet Commerce” by Z.G. Lu and H. Shen discusses privacy leakage problem in recommender systems. They presented an interesting privacy-preserving recommendation method against the kNN attacks to overcome the problem of low recommendation accuracy.

L.J. Kong and H. Shen proposed a translator trust model to transform the centralized deployment mode in S-BGP (Secure Border Gateway Protocol) to a distributed model. However, the trust (attestations of routing information) translation of TTM only depends on a single hub node and this results in severe threats for the inter-domain routing system. To overcome the deficiencies of TTM, the authors improve TTM to Distributed TTM (DTTM) by distributing the task of attestations among a set of selected multiple hub nodes instead of on a single hub node.

Parallel and distributed computing plays an important role in image processing applications that usually require a high running time. In paper "A Parallel SRM Feature Extraction Algorithm for Steganalysis Based on GPU Architecture", K. Z. Chen et al. study the well-known Spatial Rich Model (SRM) feature extraction algorithm that generates powerful Steganalysis features. But the algorithm requires to calculate a huge number of convolutions with image noise residuals, and hence has high computational complexity. They exploits parallelism of the SRM feature extraction algorithm and presents a parallel algorithm with its implementation based on GPU architecture, which significantly accelerates the extraction process.

In paper “Anomaly Detection and Localization By Diffusion Wavelet-based Analysis on Traffic Matrix”, H. Tian, T. Sun and X. Mei study Diffusion wavelets (DW)-based technique and apply them in traffic analysis. The analysis results are used in anomaly detection and localization in the paper. DW-based technique is potential to be used in many other network applications including big data analysis.

In Big Data era, how to browse and access mass information effectively from Internet, has become very hot. In paper “Topic-Sensitive Multi-document Summarization Algorithm”, N. Liu et al. propose a topic-sensitive algorithm for multi-document summarization based on Latent Dirichlet Allocation (LDA) topic. This algorithm employs not only topic features of sentences, but also statistical features of sentences.

We would like to highlight that this Special Section would not be possible without help of the organizing committee of the sixth International Symposium on Parallel Architectures, Algorithms and Programming. Also, we would like to thank Prof. Mirjana Ivanović, Editor-in-chief of ComSIS journal who gives us generous support for the special section preparation. Our special thanks also go to the editorial board for a great guidance provided while this Special Section was prepared, and to all reviewers for their valuable comments on selected papers.

Guest editor
Hui Tian
(Beijing Jiaotong University, Beijing, China)