2. Network Architecture

by Ian Pratt and Steve Hand (University of Cambridge, UK)

Link to the SOTA Chapter

Ongoing Research

Surprisingly little is known about how the Internet actually functions: One of the major ongoing challenges in the field of networking is to improve our understanding of the Internet, its traffic dynamics, trends, and the interactions between the various protocol layers. Such information is crucial to transform network management from a black art into an engineering discipline, and drive the design of future network architectures and protocols. Some of the organisations gathering such information include RIPE (see http://www.ripe.net/) and the EUDataGrid project work on 'network weather services' (see http://netmon.grnet.gr/map.shtml, for example).

Advances in photonics have led to dramatic increases in fibre-optic transmission capacity, far outstripping progress in IP router or ATM switch performance. To avoid these electronic components becoming the performance bottleneck in future networks, many groups are investigating the use of all-optical switching devices. Although optical components offer very high switching speeds, levels of integration are currently very low, hence entirely new switching techniques are required.

Future Directions

Peering forward into the future, two key challenges present themselves. The first is satisfying the demand for pervasive connectivity and seamless roaming ("Internet as infrastructure") between competing technologies and markets. The second challenge is to convert today's data networks into "human networks" with a focus on user services. Promising directions here may leverage peer-to-peer and network overlay techniques.

CaberNet Related Activities

Imperial College's area within this project is in policies for security and management of programmable (active) and adaptive networks for telecommunications.

The objective of this project is to analyse the integration of an industrial vision system in a Profibus DP Network and study the performance and dependability.

The objective of this project is to design an open architecture for the control of a diesel engine in a benchmark laboratory.

The CANU project foucuses on research topics related to the mobile ad hoc networking (MANET) characteristic of mobile devices that users carry as well as smart environments.

This project aims to extend facilities for the design of multimedia distributed systems, to ensure that they can meet the needs of complex systems which will include the use of stream communication, multicasting and Quality of Service (QoS) constraints.

The work will augment UML to support specification and design of distributed multimedia systems by providing support for streams, Quality of Service and multicasting, and provide a basis for predictions on performance and to ensure they meet QoS constraints the design environment.

The project builds on the work of the Permabase project, funded by BT Networks and Systems which designed and produced prototypes for automated performance prediction for distributed systems.

The project's goal is to develop a methodology and an architecture to improve dependability for non-safety critical, distributed, embedded automation systems with both IP (inter-site) and dedicated (intra-site) connections.

The objective of this project is to design a brake by wire system based on a Time Triggered Architecture and a bridge for a light control subsystem based on CAN network.

The objective of this project is to design a distributed architecture that uses a variation of expert systems called ruled-nets for the control of chemical process systems applied to alcoholic fermentation.

Best effort networks typically are not able to satisfy any quality of service (QoS) guarantees regarding (minimum) packet throughput, (maximum) packet delay or delay jitter. Therefore, these networks have to be modified in order to be able to support real-time communications. In this project we elaborate new techniques for fault-tolerance which allow one to accept some deficiencies in network quality (e.g. packet losses). The techniques we investigated and analyzed are either part of some dedicated middleware or they are directly supported by the distributed applications. In case of audio / video communications, which has been the focus of our studies up to now, we are analyzing, in detail, techniques such as FEC, adaptive video encoding, information dispersal, traffic smoothing as well as combinations of these techniques [Richter 2001], [Wolfinger 2001]. Both, quantitative and qualitative assessments of the improvement in video quality are focal points of our research [Heidtmann 2001].

The goal of the project is to validate by means of several fault injection techniques the communication controller of a Time triggered architecture.

The objective of this project is to develop a distributed fault tolerant architecture for a mobile robot control. This architecture uses a vision and a wireless subsystem and motion control subsystem interconnected by a fibre optics area network (CAN).

LandMARC is a Microsoft Research funded two year project at Lancaster University, UK. The main aim of the project is to build a research environment based on Microsoft technology which will support a number of important areas in mobility, distributed systems, and networking research in which Lancaster has established an international reputation.

The objective of this project is to develop a microcontroller and FPGA systems to control a laser system and vision system to take pictures of injection process inside a cylinder in a diesel engine.

The NET project will provide a emulation testbed for the performance analysis of distributed applications and protocols using a cluster system with flexible connection infrastructure.

The NETKIT project is exploring an approach to programmable networking that is inspired by the way in which middleware technology has revolutionised the (related) field of distributed applications. We believe this to be a highly promising and exciting approach that can leverage current research on specific programmable networking techniques and make these techniques significantly easier to deploy, integrate, extend and manage. In more detail, we are developing a generic, language-independent, component-based toolkit for programmable networking systems. This is building on our existing OpenCOM/ OpenORB component/ middleware technology, and is addressing all levels of programmable networking in an integrated manner from in-band packet handling on routers to high level signaling protocols. Our aim is to encompass all current programmable networking paradigms (e.g. open signaling, active networking, and application-level active networking approaches), and to support, in a uniform manner, the configuration, reconfiguration and runtime management of programmable network systems at all system levels.

Pin&play is a new approach of ad-hoc networking among objects that people can attach to large surfaces, such as notes that people pin to notice boards or artifacts that people hang on the walls in their home. It incorporates augmentation of common vertical surfaces such as walls and notice boards with low-cost conductive material to create smart surfaces as a communication medium. The objects are attached to such surfaces by means of simple pin connectors, to provide users with a familiar mechanism for adding objects to the network. This is a European IST project (project IST-2001-37007).

The R-Fieldbus project (funded by the European Commission - IST-1999-11316) aims at developing an innovative wireless communication architecture supporting industrial multimedia. Within this project the group has proposed a new hybrid wired/wireless PROFIBUS solution where most of the design options are made in order to guarantee the proper real-time behaviour of the overall network [Alves 2002]. Furthermore, the group has also participated in the development of a double stack architecture, allowing the tunnelling of TCP/IP traffic in wired/wireless PROFIBUS networks [Pacheco 2001].

The global objective of the project is to provide a characterization of the architectural abstraction levels needed to properly develop HASAs (heterogeneous access software applications), together with some proposals of analysis and verification tools, tuned to the functional and QoS properties of interest.

The SISTER project (funded by the American-Portuguese Development Foundation - FLAD 471/97) main goals were to stimulate research in the broad area of Real-Time Systems, and to support pos-graduation (mainly PhD) of the group elements. Within this project, the group developed computational models and tools to guarantee fault-tolerant real-time communication in distributed computer-controlled systems, based in industrial communication networks, such as PROFIBUS [Tovar 1999b], P-NET [Tovar 1999a] or CAN [Pinho 2000].

Organisations, particularly small and medium scale enterprises (SMEs), are finding it increasingly difficult to develop, maintain and manage their information technology (IT) applications. This is largely due to difficulties in attracting and retaining trained staff. The solution may lie in Application Service Providers (ASPs). These organisations hold the promise of providing a solution by hosting applications on their own, remotely managed servers. However, to work effectively, ASPs must guarantee security and provide resilience. Furthermore, ASPs need to ensure that hosted applications are capable of accessing a wide variety of services irrespective of the platform or the organisation through which they are provided. The mechanism for this lies in service level agreements (SLAs) over commonly available infrastructures.

  • Traffic Engineering for Service-Integrated Networks with Real-Time Communications

    • University of Hamburg, Germany

This long term project tries to approach various aspects of "Traffic Engineering" [Wolfinger 2002b] including:

  • Load measurement at different interfaces of a protocol / service hierarchy within service-integrated networks focussing on video traffic (MPEG-1/ -2, H.261, H. 263) on one hand and on UDP / IP traffic on the other hand [Zaddach 2001b];
  • Load modeling based on measurements covering again video traffic as well as packet traffic [Wolfinger 2002b];
  • Load prediction by means of modeling load transformation processes [Zaddach 2001a], [Zaddach 2002];
  • Construction of load generators for synthetical load in LANs or in Internets;
  • Traffic management, e.g. priorization of video traffic in DiffServ based networks [Ziviani 2002].


[Alves 2002] M. Alves, E. Tovar, F. Vasques, K. Roether, G. Hammer. "Real-Time Communications over Hybrid Wired/Wireless PROFIBUS-based Networks". Proceedings of the 14th Euromicro Conference on Real-Time Systems (ECRTS'02). Vienna, Austria. June 2002.

[Heidtmann 2001] K. Heidtmann, J. Kerse, T. Suchanek, B.E. Wolfinger, M. Zaddach. “Fehlertolerante Videokommunikation über verlustbehaftete Paketvermittlungsnetze“. GI-Fachtagung "Kommunikation in Verteilten Systemen" (KiVS 2001). Hamburg, Germany. Springer-Verlag. February 2001.

[Pacheco 2001] F.Pacheco, E. Tovar, A. Kalogeras, N. Pereira. "Supporting Internet Protocols in Master-Slave Fieldbus Networks". Proceedings of the 4th IFAC International Conference on Fieldbus Systems and Their Applications (FET'2001). Nancy, France. November 2001. pp. 260-266.

[Pinho 2000] L. Pinho, F. Vasques, E. Tovar. "Integrating Inacessibility in Response Time Analysis of CAN Networks". Proceedings of the 3rd IEEE International Workshop on Factory Communication Systems (WFCS'2000). Porto, Portugal. Setember 2000. pp. 77-84.

[Richter 2001] J.P.Richter. "Spezifikations- und Messmethodik für ein adaptives Dienstgütemanagement". GI-Fachtagung "Kommunikation in Verteilten Systemen" (KiVS 2001). Hamburg, Germany. Springer-Verlag. February 2001.

[Tovar 1999a] E. Tovar, F. Vasques, A. Burns. "Supporting Real-Time Distributed Computer-Controlled Systems with Multi-hop P-NET Networks". Control Engineering Practice 7(8). Pergamon, Elsevier Science. August 1999. pp. 1015-1025.

[Tovar 1999b] E. Tovar, F. Vasques."Real-Time Fieldbus Communications Using Profibus Networks". IEEE Transactions on Industrial Electronics 46(6). December 1999. pp. 1241-1251.

[Wolfinger 2001] B.E. Wolfinger, M. Zaddach. “Techniques to Improve Quality-of-Service in Video Communications via Best Effort Networks“. IEEE International Conference on Networking (ICN '01). Colmar. July 2001.

[Wolfinger 2002a] B.E. Wolfinger, P.J. Kühn (eds.). Special Issue “Internet Traffic Engineering“. Praxis der Informationsverarbeitung und Kommunikation (PIK) 24(2). June 2002.

[Wolfinger 2002b] B.E. Wolfinger, M. Zaddach, K. Heidtmann, G. Bai. “Analytical Modeling of Primary and Secondary Load as Induced by Video Applications Using UDP/IP“. Computer Communications Journal 25(11-12). July 2002. pp. 1094-1102.

[Zaddach 2001a] M. Zaddach. “Modellierung, Charakterisierung und Transformation von Videover-kehrslasten“. Dissertation, Fachbereich Informatik, Univ. Hamburg. Berichte aus dem Forschungsschwerpunkt Telekommunikation und Rechnernetze, Band 2. Wolfinger B.E. (ed.) Shaker-Verlag. Aachen, Germany. 2001.

[Zaddach 2001b] M. Zaddach, K. Heidtmann. “Measurement and Traffic Characterization of H.26x-coded Video Streams“. Procedings oft the 11th GI/ITG Conference on Measuring, Modelling and Evaluation of Computer and Communication Systems (MMB '01). Aachen, Germany. VDE-Verlag. Berlin, Germany. September 2001.

[Zaddach 2002] M. Zaddach, B.E. Wolfinger, N. Krämer, K. Heidtmann. “Lasttransformation und ihre Einsatzmöglichkeiten zur Verkehrsprognose in Intranets und im Internet“. Praxis der Informationsverarbeitung und Kommunikation (PIK) 24(2). June 2002. pp. 73-81.

[Ziviani2002] A. Ziviani, B.E. Wolfinger, J.F. de Rezende, O.C.M.B. Duarte, S. Fdida. “On the Combined Adoption of QoS Schemes to Improve the Delivery Quality of MPEG Video Streams“. 2002 International Symposium on Performance Evaluation of Computer and Telecommunication Systems (SPECTS 2002). San Diego, CA, USA. July 2002.

Maintained by Rogério de Lemos (r.delemos@ukc.ac.uk)
Last updated 4 November, 2002