Studying networks often leads to encounter complex interaction structures. These structures are often modelled using graphs. This talks is an informal review of usefull tools for researchers to gather insights on the phenomenons occurring in such structures.

More precisely, we will here discuss how and why embed graphs for visualization purposes. We will also discuss the limits of these techniques, and present other approaches to circumvent them, such as clustering analysis.

The Transmission Control Protocol (TCP) uses the well-known Slow-Start algorithm at the beginning of a connection and after idle times, which may cause needless delays. Speeding up the start of flows is one of the remaining non-trivial open issues in the Internet. Several new congestion control schemes have been developed recently in order to realize faster flow startups. Potential solutions include both end-to-end approaches and new protocols that use additional feedback from the routers on the path, such as the Quick-Start TCP extension.

This talk investigates the realization of fast startup congestion control. We introduce the fundamental challenge of resource management during flow startups and compare end-to-end and network-supported solutions. Both simulation studies and testbed measurements quantify the potential performance improvement, the robustness and fairness, as well as the benefit and cost of additional network support. The studies are performed with new implementations in the Linux network stack. Our results reveal that end-to-end fast startup mechanisms are not necessarily overly aggressive and unfair if they are selectively used. Additional signaling along the path reduces the risk of congestion at the cost of a higher complexity. Finally, several case studies prove that fast startup congestion control can indeed improve the responsiveness of broadband interactive applications.

Bio:
Michael Scharf is a research scientist at Alcatel-Lucent Bell Labs Germany. Until 2009, he was a research staff member at the Institute of Communication Networks and Computer Engineering (IKR) at the University of Stuttgart, under the supervision of Prof. Paul J. Kühn. He received his Dipl.-Ing. in electrical engineering from the University of Stuttgart in 2003. Since then, he has been an author of more than 20 conference and journal papers and an active IETF/IRTF contributor. His research interests include future Internet technologies, transport protocols, congestion control, and performance evaluation in general.

Fractional Brownian motion (fBm) emerged as a useful model for self-similar and long-range dependent Internet traffic. Approximate performance measures are known from large deviations theory for single queuing systems with fBm through traffic. We derive end-to-end performance bounds for a through flow in a network of tandem queues under fBm cross traffic. To this end, we prove a rigorous sample path envelope for fBm that complements previous approximate results. We find that both approaches agree in their outcome that overflow probabilities for fBm traffic have a Weibullian tail. We employ the sample path envelope and the concept of leftover service curves to model the remaining service after scheduling fBm cross traffic at a system. Using composition results for tandem systems from the stochastic network calculus we derive end-to-end statistical performance bounds for individual flows in networks under fBm cross traffic. We discover that these bounds grow in O(n (log n)^(1/(2-2H))) for n systems in series where H is the Hurst parameter of the fBm cross traffic and show numerical results on the impact of the variability and the correlation of fBm traffic on network performance.

Bio:
Amr Rizk is a PhD student and research assistant at the Institute for Communications Technology at the Leibniz University Hannover. He is a member of the Emmy-Noether Research Group "ProPerBounds" – a DFG funded research project on the topic "Statistical Performance Bounds for Computer networks and Communication systems". He received his Diploma in Electrical Engineering and Business Administration from the Technical University Darmstadt, Germany in 2008. His research interests are performance analysis of communication systems, queuing theory and probabilistic network calculus.

Die hohe Datenrate in modernen Netzwerken erschwert die vollständige Erfassung des Verkehrs. Die Gründe önnen sowohl in der begrenzten Performance der Hardware, als auch in der Ineffizienz von Software liegen.

Im Rahmen dieser Diplomarbeit wurden die für die Verkehrserfassung verantwortlichen Komponenten eines konventionelles Rechnersystem analysiert mit dem Ziel, die Engstellen welche die Paketverluste bei Verkehrserfassung verursachen können zu identifizieren. Dabei wurden sowohl die Hardware- als auch die Softwareaspekte betrachtet. Aufgrund der herausgestellten Problemen wurden neue Softwarekomponenten für das Betriebssystem FreeBSD implementiert. Damit wurden sowohl Paketverluste als auch die Auslastung des Rechnersystems bei der Erfassung des Verkehrs deutlich reduziert.

Mutual exclusion is a fundamental distributed coordination problem. Shared-memory mutual exclusion research focuses on local-spin algorithms and uses the remote memory references (RMRs) metric. An RMR is a shared memory access that cannot be served from a process' local cache.

A recent proof by Attiya, Hendler and Woelfel established an Ω(log N) lower bound on the number of RMRs incurred by processes as they enter and exit the critical section, matching an upper bound by Yang and Anderson. Both these bounds apply for N-process algorithms that only use read and write operations.

The lower bound of Attiya et al. only holds for deterministic algorithms, however; the question of whether randomized mutual exclusion algorithms, using reads and writes only, can achieve sub-logarithmic expected RMR complexity remained open. In this talk, I will present recent work that answers this question in the affirmative.

We present two strong-adversary randomized local-spin mutual exclusion algorithms. In both algorithms, processes incur O(log N / log log N) expected RMRs per critical section passage. Our first algorithm has sub-optimal worst-case RMR complexity of O((log N / log log N)²) . Our second algorithm is a variant of the first that can be combined with a deterministic algorithm, such as that of Anderson and Yang, to obtain O(log N) worst-case RMR complexity. The combined algorithm thus achieves sub-logarithmic expected RMR complexity while maintaining optimal worst-case RMR complexity.

The talk will be self-contained and no familiarity with Distributed Computing theory will be assumed.

This is joint work with Philipp Woelfel.

Grid Computing has, over the past few years, matured sufficiently to make it a viable solution for real-world problems. However, there are many different toolkits today that allow to build a Grid environment. The Enableling Grids for E-Science (EGEE) project is the flagship Grid project sponsored by the European Union, which brings together more than 140 institutions to produce a reliable and scalable computing resource available to the European and global research community. At present, it consists of approximately 300 sites in 55 countries and gives its 14,000 user's access to 150,000 CPU cores around-the-clock. It aims at a consolidation of existing efforts and will assist in the deployment of the resulting next generation Grid middleware (gLite) by offering support and training to new users, both in academia and industry. This session gives you an overview of what the Grid is, the architecture of the gLite middleware, an introduction to job submission, data management and grid workflows. It describes in detail the life of a job on the EGEE Grid, detailing the process of workflow management.

Bio:
Max Berger has received his Diplom at TU München, where he focused on synchronization of Personal Information Data in his master thesis. He then attended Texas Tech University, where he received his Ph.D. for designing a next-generation distributed file system. Upon his return to Europe he immediately got involved in the EGEE project, first at the Hungarian Technical Academy of Science (MTA), then at the Distributed and Parallel Systems Group at the University of Innsbruck (UIBK), where he became the local coordinator for the EGEE Project and an EGEE certified trainer. His research interest are large-scale, autonomous distributed applications.