UCL Logo Networks Research Group Logo

UCL >> Department of Computer Science >> Networks Research Group >> Seminars
People Overview Projects Publications Seminars Calls for Papers Reading Group SIGs Networking Resources PhD and Job Applications




The Networks Group meets regularly throughout the year for informal discussions, presentations, and a reading group. This is an archive of the seminar abstracts and links to the seminar slides for 2003.

August 2003

  • August 06
    Speaker:Networks Research Group
    Title: General discussion
    Abstract: Ways of combating spam and the techniques used by spammers.

  • August 13
    Speaker:Network Research Group
    Title: General Discussion
    Abstract: General Discussion

  • August 20
    Speaker:Piers O'Halon
    Title: Reading Group Paper : A Case for Network Musical Performance
    "A Network Musical Performance (NMP) occurs when a group of musicians, located at different physical locations, interact over a network to perform as they would if located in the same room. In this paper, we present a case for NMP as a practical Internet application, and describe a method to ameliorate the effect of late and lost packets on NMP. We describe an NMP system that embodies this concept, that combines serveral existing standards (MIDI, MPEG 4 Structured Audio, RTP/AVP, and SIP) with a new RTP packetization for MIDI performance. We analysze NMP experiments performed on CalREN2 hosts on the UC Berkeley, Stanford and Caltech campuses."

  • August 27
    Speaker:Rae Harbird
    Title: Reading Group Paper : A Scalable Content-addressable Network
    "Hash tables -- which map ``keys'' onto ``values'' -- are an essential building block in modern software systems. We believe a similar functionality would be equally valuable to large distributed systems. In this paper, we introduce the concept of a Content-Addressable Network (CAN) as a distributed infrastructure that provides hash table-like functionality on Internet-like scales. The CAN is scalable, fault-tolerant and completely self-organizing, and we demonstrate its scalability, robustness and low-latency properties through simulation."


September 2003

  • September 02
    TUESDAY - 11 AM
    Speaker:Michalis Faloutsos (University of California, Riverside)
    Room 203 - ROOM CHANGE
    Title:"The Internet is Like a Jellyfish" slides (ppt)
    To be advised

    September 03
    Speaker:Manish Lad
    Title: IPv6 VPN Deployment Issues slides (ppt)
    This seminar aims to outline the main issues surrounding the deployment of a dynamic IPv6-enabled VPN Infrastructure. It will summarise the main VPN technologies that UCL has examined, providing a discussion of the strengths and weaknesses of each. The current status of our investigations will be outlined along with a discussion of future aims and goals.

  • September 10
    Speaker:Christophe Diot (Intel Labs Cambridge)
    Room 203 - ROOM CHANGE
    Title: Performance of a major IP backbone network (subset of this)
    This seminar is based on the experiences gained by the speakers from Sprint's Tier-1 IP backbone. It will cover only information that is available publicly.

    The seminar starts with a brief discussion of the Internet design philosophy and the current hierarchical organization of the Internet into autonomous systems. We describe the current architecture of an IP backbone, providing a historical perspective on various aspects such as link upgrades, evolution of Points of Presence (POP), etc. This description is illustrated with examples from Sprint's IP backbone networks; elements of generalization to other backbones will also be discussed. Then we discuss routing policies and practices covering issues such as intra-domain routing, BGP configurations, relationships with other ASes and traffic engineering practices. Next, we will explain the backbone design philosophy encompassing issues such as overprovisioning, QOS, fault tolerance and manageability.

    In the second part of the seminar, we present results from an ongoing project to measure traffic on the backbone. These measures illustrate the previous discussions on backbone architecture and explain (sometimes justify) the engineering choices.

  • September 17
    Speaker:Richard Akester
    Title:ISWN: Integrated Services Wireless Network slides (ppt)
    The expression "integrated services" generally refers to the combination of data and audio/video services over the same network, which is a subject that has been studied for decades. Still, the divergent network requirements of these types of information means that no clear answers have emerged as to the best way to carry such streams. With the 802.11 wireless LAN standard appealing to both pure data users (for its replacement for wired local nets), and consumer electronics companies (for its capacity to carry multimedia), questions are again raised as to its suitability to carry both of these data types simultaneously.

    This seminar will present the speaker's experience of running DVB-T TV streams over an 802.11 network, demonstrating some of the pitfalls and unexpected problems which emerge. It will become clear that streaming high quality, high bandwidth TV over an 802.11 wireless network is itself extremely difficult, let alone mixing this with other data streams while preserving the QoS of the TV streams. Some examples of current proposals for extensions to the 802.11 standard to enable integrated services will be presented.

  • September 24 (first week of term)
    No meeting due to lack of an available room


October 2003

  • October 01
    Speaker:Tom Kelly (Cambridge)
    Title: Theoretical and implementation issues with highspeed TCP congestion control for wide area networks slides (pdf)
    Abstract: There has been a collection of proposed changes to the TCP congestioncontrol algorithm which aim to improve performance in highspeed wide area networks. The Scalable TCP algorithm design will be described and compared to other approaches. The Scalable TCP implementation on Linux will be discussed to highlight areas of current operating system practise which can be detrimental to highspeed wide area TCP performance. An experimental dataset collected from a highspeed trans-atlantic testbed will be presented.

  • October 08
    Speaker:Michael Rogers
    Title: Reading Group - Freenet uses a decentralized P2P architectureto create an uncensorable and secure global information storage system (main paper and sub-paper)
    Abstract Freenet's new routing algorithm treats response time as the sole measure of apeer's "usefulness". A request for a key is forwarded to the peer with thelowest estimated response time for that key. The estimated response time isbased on the peer's past behaviour. High-bandwidth nodes are favoured becausethe time between sending a request and receiving the end of the data ismeasured. Failure to find the requested data and failure to acceptconnections when overloaded can be factored into the estimated response timeby considering the number of past failures, and the time required to detect afailure and try another node. Thus a single quantity (derived from first-handobservations) can express a peer's reliability, locality, available bandwidthand knowledge of the network.

  • October 15
    Speaker:Tim Griffin (Intel Labs Cambridge)
    Title: Can the Border Gateway Protocol (BGP) be fixed? slides (ppt)
    Abstract: Global connectivity in the Internet is establishedand maintained by BGP. Highly expressive routing policy languagesare used to enforce connectivity relationships between independent routing domains. This policy expressiveness comes with some risk --- locally sensible policies can interact in unforeseen ways to producevarious kinds of routing anomalies such as nondeterministic routing and protocol divergence. This talk will sketch out the design space of routing policy languages and explore possible solutions to the BGP policy interaction problem.

  • October 22
    Speaker:Adam Greenhalgh
    Title: Reading Group - Zone Routing Protocol (ZRP) paper (pdf) slides (pdf)
    Abstract: Routing protocols for mobile ad-hoc networks have to face the challenge of frequently changing topology, low transmission power and asymmetric links. Both proactive and reactive routing protocols prove to be inefficient under these circumstances. The Zone Routing Protocol (ZRP) combines the advantages of the proactive and reactive approaches by maintaining an up-to-date topological map of a zone centered on each node. Within the zone, routes are immediately available. For destinations outside teh zone, ZRP employs a route discovery procedure, which can benefit from the local routing information of the zones.

  • October 29
    Speaker:Sheng Jiang
    Title: What the Grid could and should do. slides (ppt)
    Abstract: This week's seminar is a general discussion about the definition and functionality of the Grid. While the Grid has become one of the hotest topics in network field with more than half of our department's staff involved more or less with the Grid, it will be interesting and significant to have a general discussion about the definition and functionalities of the Grid. The three papers below are the recommendation reading for background on the Grid. A brief introduction about what is the Grid and the Grid problem will be given, before a discussion focusing on what the Grid could and should do. The Anatomy of the Grid: Enabling Scalable Virtual Organizations, The Grid: A New Infrastructure of 21st Century Science and The Physiology of the Grid


November 2003

  • November 05 (College Reading Week)
    Speaker:Network's Group
    Title: General discussion
    Abstract: General discussion

  • November 12
    Speaker:Lukas Kencl (Intel Labs Cambridge)
    Title: Adaptive Load Sharing for Multiprocessor Network Nodes slides (ppt)
    Abstract: This talk presents a technique for sharing the packet-processing loadamong multiple processing units within systems that act as nodes in adata communications network.

    We formulate the optimization problem of mapping packets to processors,show that the task is an NP-complete problem, and propose a heuristicmethod that uses an adaptive hash-based mapping to assign packets toprocessors. We demonstrate its advantages and prove that the methodadaptation policy possesses the key minimal disruption property withrespect to the mapping. As an example, an implementation of a serverload balancer on a network processor illustrates the applicability ofthe method.

    In the concluding part of the talk, we present some future directionsfor research in the area of adaptive methods for packet-processingsystems.

  • November 19
    Speaker:Torsten Ackemann
    Title: Reading Group - Salzer, Reed and Clark, "End-to-end arguments in system design", ACM ToCS,November 1984. paper (ps) slides (ppt)
    Abstract: This paper presents a design principle that helps guide placement of functionsamong the modules of a distributed computer system. The principle, called theend-to-end argument, suggests that functions placed at low levels of a systemmay be redundant or of little value when compared with the cost of providingthem at that low level. Examples discussed in the paper include bit errorrecovery, security using encryption, duplicate message suppression, recoveryfrom system crashes, and delivery acknowledgement. Low level mechanisms tosupport these functions are justified only as performance enhancements.

  • November 26
    Speaker:Andrew Campbell (Columbia Uni. USA)
    Title: Congestion Control in Sensor Networks
    Abstract: Sensor networks operate under light load and then suddenly become active inresponse to a detected or monitored event. This results in large, sudden andcorrelated impulses of data being generated that must be delivered to asmall set of sinks without significantly disrupting the fidelity of thesensing application. It is during these impulse periods that congestion is likely and the information being carried of most importance. We believe that without solving the congestion problem the wide-scale adoption of selforganizing sensor network technology could be jeopardized. In this talk we will discuss this problem and detail one proposal for alleviating itcalled CODA (COngestion Detection and Avoidance).

    This is work carried out with Chieh-Yih Wan and Shane B. Eisenman.
    Bio.: Andrew T. Campbell is an Associate Professor of Electrical Engineering at Columbia University and a member of the COMET Group. Andrew is working onemerging architectures and programmability for wireless networks. He received his Ph.D in Computer Science in 1996, and the NSF CAREER Award for his research in programmable mobile networking in 1999. Currently, he is on sabbatical as an EPSRC Visiting Fellow at the Computer Lab, Cambridge University.


December 2003

  • December 03
    Speaker:Network Research Group
    Title: General Discussion
    Abstract: General Discussion

  • December 10 (last week of term)
    Speaker:Mark Handley
    Title: XORP : An eXtensible Open Router Platform slides (pdf)
    Abstract: Much good Internet research is being frustrated by an inability to deploy experimental router software at points in the network where it makes most sense. This problem affects a wide range of research, including routing protocols themselves, active queue management schemes, and so-called ``middlebox'' functionality such as intrusion detection systems. Research becomes unrealistic because of the difficulty of experimenting under real network conditions; often ideas never make it out of the network simulator. Many of these problems would not exist if the router software market more closely resembled the end-system software market, which has well defined APIs for application software, and high-performance reliable open-source operating systems that permit kernel protocol experimentation.

    To address these problems, the XORP project has been developing a complete open-source router software platform, using an architecture that permits extensibility whilst at the same time providing stable well-performing basic router functionality. The vision is of an integrated router software platform running on off-the-shelf hardware, that is sufficiently well performing and reliable to see production service in a wide range of network conditions. The software architecture is designed with extensibility in mind from the start, and should permit experimental protocol support to be easily added with minimal risk of endangering pre-existing services using that router.

  • December 17
    Speaker:Mohammed Ahmed
    Title: Anarchy engenders trust! slides (ppt)
    Abstract: The pervasive environment implies a massive increase in the scale of systems, the heterogeneity of devices and diversity in services available,making the complex task of centrally managing the environment infeasible;therefore decentralised methods must be developed. The ubiquity of the environment means that devices are able to interact unbounded by physical distance and organisational boundaries; therefore insiders and outsiders cannot be easily distinguished raising the question of whom to trust? Thereby reducing the scope for trusted third parties (TTP), which assume abinary trust model with no uncertainties in mediating interaction. While the dynamic, mobile and asynchronous nature of many of the devices mean that itis difficult to predict their state, location or temporally correlate events.

    All this adds up to increased uncertainties and a need for a revision of the methods and concepts used to assess trust and provide assurance. The MARS project will address this need, in two parts: Firstly by developing modelsthat explain the effect of the physical, behavioural and organisational characteristics of components in such environments in creating sustainable decentralised trust-based mechanisms and secondly by defining more realistic models of digital trust which incorporate the uncertainties of theenvironment and are aware of the contexts of interaction in evaluating trust.