Video streaming contributes an important portion to the overall Internet traffic. Currently deployed infrastructure for video-streaming is primarily using the client-server paradigm, which is associated with high bandwidth costs for servers. To reduce bandwidth requirements on the server side, various authors suggest distributed architectures based on peer-to-peer algorithms. However, these architectures may lead to a poor quality of service for the end-user.

In this work, we combine the two paradigms and provide a framework for video-streaming providers which dynamically adapts to changing conditions, trying to maximize the benefits from both client-server and p2p organization. The proposed system will be able to handle increased load such as flash crowds by using a dynamically changing subset of clients for support, but otherwise serve requests centrally. In addition, the system enables economic usage of the underlying physical network by network-aware selection of clients for streaming support. This talk will present first ideas and planning for the next steps.

The Oracle service is an Internet-based decision facilitator that helps end-systems decide between alternative choices. It eliminates the need for end systems to infer best choices by themselves.

P2P systems can use the Oracle service to setup more efficient overlay topologies and select best sources from among multiple sources to download content from.

Building upon prior simulation results, this work aims to implement and test the performance of the Oracle service on real Internet topologies. The goal is to evaluate its efficiency and scalability in a typical P2P environment.

1st talk

Speaker: Rob Sherwood (Deutsche Telekom Laboratories Stanford)

Abstract:Despite the push for "clean slate" Internet redesign, our knowledge of the existing network and its problems is quite lacking. More specifically, because researchers have an incomplete and inaccurate view of existing networks, their ability to redesign them from scratch is hindered. In this talk, I attempt to address both points. First, I discuss my work on improving our knowledge of the Internet by producing higher quality router-level topology maps. Specifically, I describe the alias resolution problem and various techniques for solving it. Second, I describe how the OpenFlow architecture being developed at the Stanford Clean Slate lab can be applied to Internet redesign and virtual networks.

Biography: Rob Sherwood is a Senior Research Scientist at Deutsche Telekom Inc, R&D. He recently received he Ph.D. from the University of Maryland where his dissertation work focused on Internet topology discovery and network security DDoS prevention. He is currently working in concert with the Stanford Clean Slate lab on the OpenFlow and related projects.

2nd talk

Speaker: Srinivasan Seetharaman (Deutsche Telekom Laboratories Stanford)

Abstract: It is a widely accepted notion that the Internet, as it stands today, lacks the flexibility to provide enhanced services beyond basic best-effort point-to-point data delivery. As a response, "overlay networking" gained attention as a viable alternative to overcome this network ossification. Overlay networks (such as Service overlay networks, P2P systems) implement complex functionality without modifying the underlying IP routers, by forming an independent and customizable virtual network over the native IP network. This approach, however, can introduce complex "cross-layer interactions" between the native and overlay layers. In this talk, I will present a brief summary of my earlier analysis of cross-layer interaction in overlay networks. I will follow it up with short term research plans in the context of a service-oriented future Internet. Specifically, I will talk about some exciting projects we plan to undertake in the Clean Slate group to enhance services offered to end-users.

Biography: Srinivasan Seetharaman is a Senior Research Scientist in Deutsche Telekom Inc R&D. He received the Ph.D. degree in Computer Science from the Georgia Institute of Technology, and a Masters degree in Computer Science from The Ohio State University. His research interests lie in the areas of large-scale networked systems, with emphasis on network architecture, performance evaluation, network monitoring and networking protocols.

One of the major reasons for the rising demand of high bandwidths in DSL broadband environments is the distribution of video content. This includes not only content providers like youtube.com or video.yahoo.com, but also TV channels originally broadcast via terrestrial, cable or satellite and now being distributed by packet switched networks. While the former are transmitted without any methods assuring service quality, a great variety of complex technical approaches are necessary to provide failure-free delivery of multimedia content over best-effort networks. Even though packet losses can not be completely avoided, especially for the last mile to the customer, network providers already deployed several different mechanisms like prioritizing data streams, forward error correction (FEC) or retransmissions to keep losses at a minimum.

Deutsche Telekom Laboratories' project T-V-Model, in which context this thesis was prepared, aims to establish a model to describe the IPTV service quality perceived by a user (Quality of Experience—QoE). Up to now, error correction methods like retransmissions are not included. The present thesis aims to extend this framework by providing a model which describes the transformation of a lossy input data stream to the resulting packet loss process after a retransmission scheme is applied. Statistical properties of the loss process can be replicated by a model, which can also be used to generate synthetic loss traces of arbitrary length for the study of QoE impacts.

Measurements were conducted to discover and model the set-top-box's (STB) characteristic reactions on impaired data streams. This results in a state model describing the typical behavior of a STB, which has been used in further simulations involving packet losses caused by phases of overload. The latter study revealed a new approach that allows loss pattern without retransmission to be mapped to loss pattern with retransmission. Statistical properties of the studied loss pattern could be replicated by a modified Gilbert-Elliott Markov model, which leads to new opportunities in T-V-Model's Quality of Experience measurements.

Routing tables in the Default Free Zone (DFZ) of the Internet are steadily growing and becoming harder and harder to manage and maintain. The size of the tables and the frequency with which they need to be updated is a huge snag to deploy enhanced services, e.g. mobility or fine-grained traffic engineering.

We propose a new routing architecture which exploits flat identifiers and hierarchical locators. We maintain the binding between identifiers and locators in a hybrid push/pull mapping system which is designed to be scalable and to keep updates localized. Since traffic engineering, multihoming and mobility are natively supported, our approach also makes BGP useless for a lot of access networks. In turn, this reduces the number of destinations that interdomain routing has to deal with.

In comparison with other approaches, our solution pushes state and resource-consuming tasks to the end hosts, relieving core routers from unneeded workload. We discuss the advantage of our solution and demonstrate its incremental deployability via a prototype implementation.

This is joint work with Randy Bush, Anja Feldmann, Olaf Maennel, and Wolfgang Mühlbauer.

In Next Generation Networks (NGNs), vertical handovers between different network technologies provide seamless roaming during active VoIP sessions. The quality of these VoIP sessions, as well as quality of user experience, depends on the network handover policies, link layer characteristics, and speech codecs used in the respective networks. Even though extensive research has been carried out on seamless mobility, a thorough analysis of user perception of this phenomenon is still missing in order to successfully design and further improve mobility management solutions for always-on multimedia services. During this talk we will present an NGN testbed and a project methodology applied to map user experience to network conditions, focusing on phenomena caused by a user roaming across diverse wireless technologies. The results present the speech quality as a function of provided audio bandwidth and network characteristics, as well as of the temporal handover position within a call. The evaluation of these test results enables us to specify circumstances when a high-quality network handover should be scheduled. This is important for the development of always-on multimedia services that provide maximal user satisfaction.

Bio
Blazej Lewcio joined Deutsche Telekom Laboratories in 2007 and works towards his PhD on Quality of Service and Quality of Experience in Next Generation Networks, which includes Networking and Usability aspects. He graduated from Deutsche Telekom Hochschule für Telekommunikation in Leipzig, Germany (University of Applied Science). During his study, he focused on computer networks and distributed systems. He completed his study with a diploma thesis on Assessment and adaptation of voice quality in All-IP heterogeneous networks, which he wrote as a member of the Mobisense project under supervision of Dr. Pablo Vidales.

Over the last decade, unsolicited bulk email, or spam, has transitioned from a minor nuisance to a major scourge, adversely affecting virtually every Internet user. While there is a considerable body of research focused on spam from the recipient's point of view, we understand considerably less about the sender's perspective: how spammers test, target, distribute and deliver a large spam campaign in practice.

In this talk, I will report on our exploration of a new methodology – distribution infiltration – for measuring spam campaigns from the inside. By hooking into a botnet's command-and-control (C&C) protocol, one can infiltrate a spammer's distribution platform and measure various aspects of spam campaigns as they occur. I will present an initial analysis of spam campaigns conducted by the Storm botnet, based on data we captured by infiltrating its distribution platform, including such aspects as address harvesting, the template language employed, and dynamic aspects such as target list and campaign sizes.

Video streaming with virtual pan/tilt/zoom functionality allows the viewer to watch arbitrary regions of a high-spatial-resolution scene. In our proposed system, the user controls his region-of-interest (ROI) interactively during the streaming session. The relevant portion of the scene is rendered on his screen immediately. An additional thumbnail overview aids his navigation. I will present a video coding scheme that efficiently supports random access to arbitrary ROIs, while keeping the transmission rate as low as possible. Following this, I will present a peer-to-peer (P2P) multicast live video streaming system to provide the control of interactive region-of-interest (IROI) to large populations of viewers while exploiting the overlap of ROIs for efficient and scalable delivery. The proposed P2P overlay is altered on-the-fly in a distributed manner with the changing ROIs of the peers. The main challenges for such a system are posed by the stringent latency constraint, the churn in the ROIs of peers and the limited bandwidth at the server hosting the IROI video session. Experimental results with a network simulator indicate that the delivered quality is close to that of an alternative traditional unicast client-server delivery mechanism yet requiring less uplink capacity at the server.

Biography:
Aditya Mavlankar is currently a researcher at the Information Systems Laboratory of Stanford University, Stanford, CA. He got his B.E. degree in Electronics and Telecommunication from University of Pune, India in 2002 and his Master's degree in Communications Engineering from the Technical University of Munich, Germany in 2004. His Master's thesis was awarded the Edison Prize Bronze Medal by IIE Europe in conjunction with the GE Foundation. He has been twice the co-recipient of the Best Student Paper Award; at the IEEE Workshop on Multimedia Signal Processing (MMSP), Victoria, Canada in 2006 and at the 15th European Signal Processing Conference (EUSIPCO), Poznan, Poland in 2007. His paper at the 16th International Packet Video Workshop (PV) in Lausanne, Switzerland in 2007 won him a Student Travel Grant Award. He co-authored a paper at the Picture Coding Symposium (PCS) in Lisbon, Portugal in 2007 which was among the top four Finalists for the Best Paper Award. His research interests include scalable video coding, interactive video delivery and peer-to-peer video streaming.

In many situations it can be enormously helpful to archive the raw contents of a network traffic stream to disk, to enable later inspection of activity that becomes interesting only in retrospect. We present a Time Machine (TM) for network traffic that provides such a capability. The TM leverages the heavy-tailed nature of network flows to capture nearly all of the likely-interesting traffic while storing only a small fraction of the total volume. Our system provides flexible, high-performance traffic stream capture, indexing and retrieval, including an interface between the TM and a real-time network intrusion detection system (NIDS). The NIDS controls the TM by dynamically adjusting recording parameters, instructing it to permanently store suspicious activity for offline forensics, and fetching traffic from the past for retrospective analysis. We present a detailed performance evaluation of both stand-alone and joint setups, and report on experiences with running the system live in high-volume environments.

This is joint work with Robin Sommer, Holger Dreger, Anja Feldmann, Vern Paxson, and Fabian Schneider and will be presented at SIGCOMM 2008 in Seattle, WA.

A significant portion of the Internet Traffic nowadays is contributed by Peer-to-Peer systems and a major problem is represented by polluted files, which implicates that network bandwidth is wasted and the user experience is depleted.

Existing solutions for reducing pollution are only based on reputation of the peer but don't take peer locality into account, so network bandwidth is still being wasted. These solutions are also susceptible to unfair rating, traitors, collusion, front peers and Sybil attacks.

A new solution by Aggarwal, et.al. proposes that the ISP can offer an oracle service to the P2P users, which gets from the user a list of possible neighbors (their IP addresses) which is then sorted according to some proximity criteria decided by the ISP, e.g., network locality, last-hop bandwidth, latency, routing policy, etc.

This thesis evaluates if trust can help to reduce pollution in P2P systems with help of the oracle, without degrading the proximity values. The goal of this thesis is only to show whether trust can be used in P2P systems. Therefore we use a basic trust system.

Several simulations have been done using the application level simulator PeerSim. We show the results of different percentage combinations of trust and proximity and different distribution of P2P nodes within the ASes.

We find that already a small trust percentage helps to reduce pollution drastically, from 50% to 10%, without degrading the proximity values significantly.

We also have done experiments on Multiple-ISP cooperation and showed that it is possible to find the overall best peer for downloading if oracles cooperated with each other.
Thus the oracle service makes it possible to download content in short time and from a near and trustworthy peer.

We evaluate the performance of a large-scale live P2P video multicast session comprising more than 120,000 peers on the Internet. Our analysis highlights P2P video multicast characteristics such as high bandwidth requirements, high peer churn, low peer persistence in the P2P multicast system, significant variance in the media stream quality delivered to peers, relatively large channel start times, and flash crowd effects of popular video content. Our analysis also indicates that peers are widely spread across the IP address space, spanning dozens of countries and hundreds of ISPs and Internet ASes. As part of the P2P multicast evaluation several QoS measures such as fraction of stream blocks correctly received, number of consecutive stream blocks lost, and channel startup time across peers. We correlate the observed quality with the underlying network and with peer behavior, suggesting several avenues for optimization and research in P2P video multicast systems.

Venue:
IEEE 16th International Workshop on Quality of Service (IWQoS 2008), University of Twente, Enschede, The Netherlands

Paketerfassung in Hochgeschwindigkeitsnetzen stößt an verschiedenen Grenzen an. In diesem Vortrag werden die Grenzen (Speicher/Bus-Bandbreite, Interrupt-last, …) aufgezeigt und diskutiert. Um diese Beschränkungen zu beseitigen, werden verschiedene Konzepte vorgestellt, von denen mind. eines im Rahmen der DA umgesetzt werden soll.

Talk will be held in German.

Unexpected surges in Web request traffic can exercise server-side resources (e.g., access bandwidth, processing, storage etc.) in undesirable ways. Administrators today do not have requisite tools to understand the impact of such flash crowds on the their servers. Most Web servers either rely on over-provisioning and admission control, or use potentially expensive solutions like CDNs, to ensure high availability in the face of flash crowds. A more fine-grained understanding of the performance of individual server resources under emulated but realistic and controlled flash crowd-like conditions can aid administrators to make more efficient resource management decisions. We present mini-flash crowds (MFC) – a light-weight profiling service that reveals resource bottlenecks in a Web server infrastructure. MFC uses a set of controlled probes where an increasing number of distributed clients make synchronized requests that exercise specific resources or portions of a remote Web server. We carried out controlled lab-based tests and experiments in collaboration with operators of production servers. We show that our approach can faithfully track the impact of request loads on different server resources and provide useful insights to server operators on the constraints of different components of their infrastructure. We also present results from a measurement study of the provisioning of several hundred popular Web servers, a few hundred Web servers of startup companies, and about hundred phishing servers.

This is joint work with

• P. Ramamurthy, A. Akella (UW-Madison), V. Sekar (CMU), A Shaikh (IBM Research)

and will be presented at the USENIX Annual Technical Conference, June 2008

http://www.research.att.com/~bala/papers

We examine how information related to individual users is aggregated as a result of browsing seemingly unrelated Web sites. The privacy diffusion through a diversity of hidden transactions is widespread. The potential for a few third party aggregators to construct a profile of individual user's movements on the Web is high.

We study thousands of Web sites across numerous categories, countries, and languages. We generate a privacy footprint allowing us to assess and compare the diffusion of privacy information across a wide variety of sites. We examine the effectiveness of existing and new privacy protection techniques to reduce this diffusion. We also explore the impact on the quality of Web pages due to protection techniques. Using objective measures for privacy protection and page quality, we examine their tradeoffs for different privacy protection techniques applied to a collection of popular Web sites as well as a focused set of sites with significant privacy concerns.

We also take a quick look at privacy issues on Online Social Networks where users often willingly, share personal identifying information about themselves without a clear idea of who accesses it. We examined popular OSNs and identify what bits of information are being shared, how widely, and what users can do to prevent such sharing. We also examine the role of third-party sites that track OSN users and compare with privacy leakage on popular traditional Web sites.

http://www.research.att.com/~bala/papers

Our group is actively involved in EU FP-7 projects and has ties to related activities. However, for many of us, acronyms like Carmen, Trilogy and 4WARD are either unknown or, well, just acronyms.

The goal of this talk is to ascertain a common and basic understanding of the projects we are involved and in which context they are executed.

Towards this goal, I initially and briefly talk about the EU FP-7 program in general: what is an EU project, what does it contain, and why are we involved? Examples come from the Carmen project.

Then, the main talk focuses on the Clean Slate Internet Design efforts. The objectives of Trilogy are explained and put into context with related projects from FP-7 and NeTS/FIND.

Real-time Internet services are gaining in popularity due to rapid provisioning of broadband access technologies. Delivery of high Quality of Experience (QoE) is important for consumer acceptance of multimedia applications.

IP packet level errors affect QoE and the resulting quality degradations have to be taken into account in network operation. This talk shows how second order statistics of the number of packet losses in finite Markov models over several relevant time-scales can be derived and used to adapt models to the loss process visible in traffic measurements in wired and wireless networks.

Markov models obtained in this way provide a generator for packet loss pattern to be used in the estimation of the degradation in the Quality of Experience (QoE) for Internet services, especially real-time video streams.

The evaluation of statistical properties of video streams impaired by Markov models generating packet loss shows shows different error pattern at the application (video) layer depending on the used parameter estimation method. This finding highlights the need for models replicating error pattern which are relevant for higher layers.

Many routing protocols in mobile ad hoc networks use a flooding mechanism to allow a source node to find its destination node. Flooding the network means broadcasting a route discovery message to each node in the network, in the hope that a node with an active route to the destination will reply with the route reply. It is a well known fact that flooding impacts the performance of the network negatively. On a shared wireless medium, when all nodes transmit at the same time to forward the route discovery packets, they create contention, congestion, and race conditions.

In this presentation, we will consider three different facets of the impact of flooding, based on some recent work:

• we will present an analytical framework to quantify the cost benefit of flooding the network in terms of discovering the network topology. In particular, our analysis allows to predict the gain of sending a route discovery for protocols like AODV and DSR. Using the analytical framework, we will show that there is a sweet spot where flooding makes sense.

• we will present a protocol which attempts to minimize the flooding in the network while taking advantage of the sweet spot identified earlier. This protocol is based on AODV and improves the performance while decreasing the overhead.

• finally, we will study the performance of the so called bread crumbs protocol, which allows to locate a node without resorting to flooding. We will show that the path created using this protocol is asymptotically equivalent to the shortest path, up to a constant factor.

Biography:
Cedric Westphal received his PhD in Electrical Engineering from UCLA in 2000. He has been a visiting researcher at Stanford University from 1997–2000, and a visiting professor at Keio University, Japan, in 2006–2007. He has worked at Nokia Research, Nokia-Siemens-Networks and is currently with DoCoMo Labs USA. He has published extensively in the area of ad hoc networks and wireless networks. He is on the editorial board for Computer Networks journal, and has participated in the organization and the program committee of many ACM and IEEE conferences, including Infocom, MobiCom and MobiHoc.

Comparison of graph structures is a frequently encountered problem across a number of problem domains. To perform a useful comparison requires definition of a cost function that encodes which features of the graphs are considered important. Although the spectrum of a graph is often claimed to be a way to encode a graph's features, the raw spectrum contains too much noise to be useful on its own. In this paper we have introduced a new cost function, the weighted graph spectrum, that improves on the graph spectrum by discounting those eigenvalues that are believed to be unimportant and emphasizing the contribution of those believed to be important.

We use this cost function to optimise selection of parameter values within the particular problem domain of Internet topology generation. The weighted spectrum was shown to be a useful cost function in that it leads to parameter choices that appear sensible given prior knowledge of the problem domain (close to the default values and, in the case of the BA generator, falling within the expected region). In addition, as the metric is formed from a summation, it is possible to go further and identify which particular eigenvalues are responsible for significant differences.

We are in the early years of a revolution in wireless communications and the future holds many possibilities. Central to this revolution is the ability to dynamically program (and reprogram) radios, so that innovation in radios will come at the speed of software releases rather than hardware releases. But programmability is only part of the story. There are regulatory issues. There are potential innovations in energy use. In this talk I sketch how how research innovations in wireless are likely to unfold over the next dozen years or so with some (naive) discussions of how the innovations will affect the marketplace.

Dr. Craig Partridge is Chief Scientist for Networking Research at BBN Technologies, where he has done networking research since 1983. Craig is a former chair of ACM SIGCOMM and former editor-in-chief of IEEE Network Magazine and ACM Computer Communication Review. He is best known for designing how email is routed (MX resource records) and for leading the effort to build multigigabit routers in the mid-1990s. He is both an IEEE and ACM fellow and holds his A.B., M.Sc. and Ph.D. degrees from Harvard University.

The Global Environment for Network Innovation (GENI) is a US national-scale facility for conducting networking and computer science research, now in its design and planning phase. It will support a future, experimental network infrastructure that will allow researchers from diverse disciplines across computer and information science and engineering, as well as from economics and the social sciences, to escape today's Internet-circumscribed research environment. This infrastructure will allow researchers to program their own unique networks (i.e., multiple end-to-end "slices") that run on federated, heterogeneous backbones and edge networks to try out new protocols (that may or may not be TCP/IP-based), global-scale "clean slate" architectures, and/or new cross-layer research whereby mechanisms that incorporate human values, such as information privacy or security, for example, become integral parts of the network architecture as opposed to an afterthought. The intention is for this infrastructure to be deeply instrumented so that researchers will be able to monitor their ongoing experiments, collect data on the use of novel protocols, designs and architectures, and analyze the emergent behaviors of traffic.

In this talk, I'll present an overview of GENI, the various groups playing a role in planning GENI, and how we expect GENI to be developed and to evolve over its lifetime.

Dr. Craig Partridge is Outreach Director for the GENI Project Office and Chief Scientist for Networking Research at BBN Technologies, where he has done networking research since 1983. Craig is a former chair of ACM SIGCOMM and former editor-in-chief of IEEE Network Magazine and ACM Computer Communication Review. He is best known for designing how email is routed (MX resource records) and for leading the effort to build multigigabit routers in the mid-1990s. He is both an IEEE and ACM fellow and holds his A.B., M.Sc. and Ph.D. degrees from Harvard University.

As Magnets got its first nodes running, it is time to think about how to proceed. I'd like to discuss some thoughts of mine with all of you:

• Status update: what is Magnets doing.

• Infrastructure: We have a lot of things to do for our infrastructure. Which ones are worth doing? How to partition them for student projects? We'll get an admin at some point, how to make use of that …

• Research: My proposal for an mesh-to-internet gateway. What would you do differently? What to do now, for getting things comparable with recent simulated results later …

• Roger and Thomas will also add some topics.

I hope for very informal PGT, with few slides and a lot of discussion. (even for PGT standards)