A Methodology for Estimating Interdomain Web Traffic
Speaker: 
Prof. Dr. Bruce MacDowell Maggs
Carnegie Mellon University & Vice President, Research, Akamai Technologies
This talk introduces a new methodology for measuring interdomain web traffic flows, and how they change over time. We are able to estimate, for each of over one thousand web sites, including approximately fifty of the one hundred most popular web sites in the United States, the volume of traffic to and from all clients worldwide. Our technique begins by examining the logs of the thousands of reverse proxy servers deployed by Akamai Technologies to serve content on behalf of Akamai's customers, the web sites. These logs give a complete view of all traffic served by Akamai to clients around the world. We then extrapolate from the Akamai data to estimate the total traffic served directly by the web sites to the clients. In particular, through passive monitoring of all web traffic in and out of several universities, we determine the ratio of bytes served by each web site to bytes served by Akamai. This is a significantly larger volume of traffic, and, to a greater extent than the Akamai traffic, travels across multiple domains. The talk concludes with a preliminary examination of the data.
Joint work with Anja Feldmann, Nils Kammenhuber, Olaf Maennel, Roberto De Prisco, and Ravi Sundaram.
The Role and Impact of Distribution in Software and Systems Engineering
Speaker: Prof. Dr. Dr h.c. Manfred Broy
Today, in general, software is multi-functional, mutually connected, embedded, distributed over networks and structured into logical components that interact asynchronously with sensors, actuators, human machine interfaces and networks. The distribution of such systems is one of their dominant characteristics. Their development requires advanced models to master their complexity.
We examine fundamental models of distributed software systems and their properties, identify and describe various basic views, and show how they are related. We concentrate on models of distributed composed systems that interact by message exchange. We consider, in particular, models of data, state, interface, functionality, distribution, hierarchy of composed distributed systems, and process. We study relationships between models and views by abstraction and refinement as well as forms of composition and modularity. In particular, we introduce a comprehensive mathematical model and algebra for composed distributed systems, essential views and relationships.
Large Networks and the Price of Anarchy or Stability
Speaker: Prof. Dr. Eva Tardos
Cornell University, Ithaca
The emergence of the Internet and the Web represents a radical shift of focus in our thinking about computational systems. Perhaps the most important distinguishing feature of these networks is that they are simultaneously built, operated, and used by multiple parties with diverse sets of interests and with constantly changing degrees of cooperation and competition. One of the main challenges faced by computer science today is to successfully build and manage such systems.
Networks that operate and evolve through interactions of large numbers of participants play a fundamental role in many domains, ranging from communication networks to social networks. They can also naturally model the behavior of many physical systems, such as electricity in electric networks, and the distribution of forces in mechanical structures. In this talk we will consider some simple network games modeling routing or network design where multiple agents each pursue their own selfish interests. Our goal is to quantify the degradation of quality of solution caused by the selfish behavior of users, compared to a centrally designed optimum.
Algorithmic issues in Coalitional and Dynamic Network Games
Speaker: Prof. Dr. Paul G. Spirakis
University of Patras & Director of the Computer Technology Institute
We discuss here some new algorithmic and complexity issues in coalitional and dynamic/evolutionary games, related to the understanding of modern selfish and Complex networks.
In particular:
- We examine the achievment of equilibria via natural distributed and greedy approaches in networks.
- We present a model of a Coalitional game in order to capture the anarchy cost in such situations.
- We propose a stochastic approach to some kinds of local interactions in networks , that can be viewed also as extensions of the classical evolutionary game theoretic setting.
The SFB 376 "Massively Parallel Computing": development and some recent results
Speaker: Prof. Dr. Friedhelm Meyer auf der Heide
The Collaborative Research Center (Sonderforschungsbereich, SFB) began working in 1995 with the aim of developing
- algorithmic methods and techniques for exploiting the computational capabilities of parallel processor networks and
- design methods for embedded, distributed (technical) systems.
The quality of the algorithmic solutions and the design methods should be evaluated in applications of high scientific and/or economic interest.
Since then, the SFB has made essential contributions to this topic area and acquired international reputation. In addition, it takes account of the rapidly growing importance of networks as communication and information systems as mentioned above. Apart from the aspect of parallel high-performance computers, research has therefore begun to focus increasingly on areas such as usability of heterogeneous dynamic networks, for example mobile, wireless communication networks.
In the talk, I will survey the development of the SFB, and focus on recent work on data management in dynamic networks.
Self-Coordinating Systems: The Next Challenge in Research on Distributed Systems
Speaker: Prof. Dr. Franz Josef Rammig
Most technological artifacts will become intelligent "things that think" and most of these intelligent objects will be linked together to an "Internet of things". To master this omnipresent virtual "organism", completely new design and operation paradigms have to evolve. In this paper, we present our view on this challenging research area. We structure it into theoretical foundations, characterized as Coordination Paradigms, and three major application-driven fields - Self-coordinating Objects, Self-coordinating Networked Objects, and Self-coordinating Anthropomatics. Major research questions and potential approaches are discussed. Finally, it is pointed out that CRC376 "Massively Parallel Computing" has already made initial contributions to this highly demanding research area.
