Computer Networks 52 (2008) 2817–2818
Contents lists available at ScienceDirect
Computer Networks journal homepage: www.elsevier.com/locate/comnet
Guest Editorial
Complex computer and communication networks
Complex network structures, generally modeled as large graphs, have played an important role in recent computer network research. Typical examples include physical connectivity structures such as the Internet’s router-level topology; more logical or virtual maps such as the Internet’s AS-level (Autonomous System-level) graphs; overlay networks such as the Web graph or peer-to-peer systems; social networks like email exchange graphs or online contacts; and sensor and/or mobile networks. Recent advances in complex network research have significantly improved our understanding of general networked systems. However, the results of this work are not necessarily applicable to real-world computer and communication networks, where nodes and links have system-specific meaning and where connectivity properties cannot in general be dealt with in isolation, but have to be viewed within the context of the traffic that traverses these networks and of the dominant protocols that determine how this traffic flows across them. Owing to these new challenges, the study of complex computer and communication networks has become a very active area of research and includes aspects such as measurement, analysis, modeling, and algorithms. 1. Measurement Measurements are critical for most of the papers in this Special Issue. They are either used directly to evaluate an algorithm or protocol or indirectly to develop models and construct graph structures for simulations. However, many real-world complex networks are not directly accessible, and in many cases, one has to perform carefully designed measurement studies to obtain and collect the desired information about these networks. The resulting measurements typically provide only partial views of the networks, and in addition these views may be inaccurate or biased in the sense that the observed properties may be the result of the measurement procedures themselves and may say little about the underlying actual network. The last few years have seen an increasing interest in designing and applying more accurate measurement methods for various complex networks of interest. Contri1389-1286/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.comnet.2008.06.001
butions in this area include novel techniques for detecting, estimating, and correcting for measurement bias; new methods for sampling large-scale and highly heterogeneous graphs; and network tomography approaches. The paper by M. Fraiwan and G. Manimaran titled ‘‘Scheduling algorithms for conducting conflicting measurements in overlay networks” contributes to this area by addressing the measurement conflict problem associated with active network measurements. 2. Analysis A main goal of analyzing complex networks is to extract interesting information and illuminating properties from them. To achieve this objective, the traditional approach is to focus on static graph structures and consider a number of different graph metrics or statistics, including size, density, average path length, diameter, degree distribution, and clustering coefficient. Higher-order statistics that have been studied are degree correlations (for total, in-, or outdegree), correlations between node degree and clustering coefficient, betweenness centrality of nodes and links, connectivity structures, community structures, etc. More recently, there has been interest in analyzing graph structures that evolve over time, but the analysis of such dynamic graphs is still in its infancy. This Special Issue contains two papers that are concerned with analyzing complex networks. Although somewhat more algorithmic in nature, the paper by M. Gonen et al. titled ‘‘Finding a dense-core in jellyfish graphs” studies certain types of static graphs and aims at finding the dense-core in such graphs. In contrast, the paper by A. Scherrer et al. on ‘‘Description and simulation of dynamic mobility networks” focuses on evolving networks and presents a framework for analyzing dynamic mobility networks based on sensor measurements. 3. Modeling Modeling is a main aspect of complex network research. The quality of proposed network models is traditionally judged by how well they fit the underlying data; that is,
2818
Guest Editorial / Computer Networks 52 (2008) 2817–2818
capture their main properties. Other criteria include simplicity, parsimony, and predictive power. Network models are used for a variety of purposes, including the generation of synthetic network structures for simulations, prediction of the evolutionary behavior of the network, and gaining an understanding of the key forces that shape the structure and impact the evolution of the network. Such ‘‘evocative” models are generally more difficult to derive than the more conventional ‘‘descriptive” models, but their potential to reverse-engineer observed complex networks makes them very attractive for studying complex computer and communication networks. A number of papers in this Special Issue include a significant modeling component. For example, simple models that allow the generation of synthetic dynamic mobility graphs are proposed in the paper by A. Scherrer et al. A different aspect of modeling is considered in the paper ‘‘Policy relationship annotations of predefined AS-level topologies” by A. Vilhar and R. Novak, who are concerned with realistic labeling of the edges in AS-graphs of the Internet.
In summary, this Special Issue contains seven papers that are representative of the current state-of-the-art in the area of complex computer and communication network research. In addition to making significant contributions to the field, these papers also illustrate the breadth of problems and scientific challenges and the need for innovative approaches and creative solutions faced by researchers who are interested in working in this area. Matthieu Latapy Universite Pierre et Marie Curie, CNRS, Laboratoire LIP6, 104 Avenue du President Kennedy, 75016 Paris, France Tel.: +33 1 44 27 87 84; fax: +33 1 44 27 74 95 E-mail address:
[email protected] Walter Willinger AT&T Labs, 180 Park Avenue, Building 103, Florham Park, NJ 07932, USA E-mail address:
[email protected] Available online 8 June 2008
4. Algorithms Dealing with large-scale graphs naturally poses many algorithmic issues. In particular, the design, management, and control of large complex networks as well as their analysis give rise to new algorithmic problems which are largely non-standard from the point of view of classical graph algorithms. These problems require new approaches, inspire the development of novel algorithms, and challenge conventional evaluation methodologies (both formal and experimental). For example, given that many real-world complex networks are believed to share some important common properties, it may be advantageous to design new graph algorithms that exploit some of these properties. Such algorithms are likely to be somewhat inefficient when applied to general graphs, but they may be highly efficient when applied to certain types of real-world complex networks. Algorithmic issues are at the heart of all the papers in this Special Issue. A typical example is the paper by K.-H. Vik et al. on ‘‘Evaluating Steiner tree heuristics and diameter variations for application layer multicast”, where the authors devise new and efficient Steiner-tree constructions for multicast communication. A different algorithmic problem is considered in the paper by I. Gojmerac et al. titled ‘‘Towards low-complexity Internet traffic engineering: The adaptive multi-path algorithm” that describes and evaluates a new algorithm for distributing load within a network domain by off-loading congested links in realtime. Finally, the paper by J. Karlin et al. on ‘‘Autonomous security for autonomous systems” describes, analyzes, and evaluates a new BGP-like routing protocol that ensures a level of security without enforcing centralized control in a distributed system.
Matthieu Latapy (
[email protected]) received his MS degree in Computer Science and his PhD in Algorithmics from University Paris Diderot (Paris 7), France. After a postdoc at INRIA, he obtained a CNRS permanent researcher position. He first joined the LIAFA, CNRS and University Paris Diderot, and later moved to LIP6, CNRS and University Pierre and Marie Curie (Paris 6), France. He leads the LIP6 team on Complex Networks (http:// complexnetworks.fr). The research conducted in this group is focused on real-world complex networks modeled as graphs, like internet topologies, web graphs, exchange graphs, or social networks. It has a strong interdisciplinary nature, dealing with both different topics in computer science (algorithmics, networking, visualization, ...) and different fields (computer science, mathematics, physics, social sciences, ...). Another original characteristic of this group is that it addresses all kinds of problems raised by real-world complex networks, from measurements to algorithmics, and including modeling and analysis.
Walter Willinger (walter@ research.att.com) received the Diplom (Dipl. Math.) from the ETH Zurich, Switzerland, and the M.S. and Ph.D. degrees from the School of ORIE, Cornell University, Ithaca, NY. He is currently a member of the Information and Software Systems Research Center at AT&T Labs Research, Florham Park, NJ, and before that, he was a Member of Technical Staff at Bellcore Applied Research (1986–1996). He is a Fellow of ACM (2005), a Fellow of IEEE (2005), and was named an AT&T Fellow in 2007. For his work on the self-similar (‘‘fractal”) nature of Internet traffic, he received the 1996 IEEE W.R.G. Baker Prize Award from the IEEE Board of Directors, the 1994 W.R. Bennett Prize Paper Award from the IEEE Communications Society, and the 2006 ACM SIGCOMM ‘‘Test of Time” Paper Award.