Application level active networking

doi:10.1016/S0169-7552(98)00293-1

Computer Networks

Volume 31, Issue 7, 8 April 1999, Pages 655-667

https://doi.org/10.1016/S0169-7552(98)00293-1 Get rights and content

Abstract

In this paper we describe and discuss an Application Level Active Network system. This system provides the benefits of proposed Active Networks, including rapid and transparent deployment of new network services. However our system is also relatively free of the problems of router-level Active Network deployment, such as concerns over safety and resource management. We describe our overall architecture and its components. We then describe and discuss an implementation of the architecture in Java. We present a number of applications that have been implemented on the architecture, and indicate the benefits of our approach.

Introduction

Currently the deployment of new communication services is limited by the slowness of standardisation processes and the inflexibility of the communications infrastructure. Recently an approach to overcoming these problems has been proposed in the form of Active Networks (AN) [11]. AN researchers are currently proposing the deployment of protocol elements at the network router level. We believe this approach to be unrealistic. It is most unlikely that a network provider will permit the deployment of protocol code from third parties at this level. This would introduce intolerable security and safety issues, and could have a serious impact on the level of service experienced by the multiple network flows sharing the router.

We propose an Application Level Active Network (ALAN) system. Such a system consists of regular clients and servers, such as WWW [4]browsers and servers, located on the Internet or Intranet. Communication between servers and clients is enhanced by Dynamic Proxy Servers (DPS) that are located at optimal points of the end-to-end path between the server and the client. There may be more than one DPS involved in a end-to-end path. It is possible to download protocol entities onto the DPS infrastructure. These protocol entities then act as filters or enhanced protocol functionalities that improve the level of service between servers and clients. Protocol modules may be obtained from protocol servers, or more general servers such as Web servers, owned by network operators or value added service providers.

The idea of special proxy machines has a well understood parallel in the current WWW caching infrastructure. Dynamic Proxy Servers, and the associated proxylet and protocol servers, will typically be owned by a network provider or the owner of a private Intranet. This constraint, we believe, makes security issues more tractable. Acceptable security can be achieved using known mechanisms.

In this paper we firstly describe our overall ALAN architecture and its components, using a typical application scenario. We then describe and discuss how the architecture and its components are implemented in Java. A successful implementation of an audio streaming example indicates that the architecture is realisable. We describe further application examples that illustrate generality, and present some results that demonstrate the benefits of our approach. We conclude by discussing future issues.

The main contribution of our approach is to enable rapid deployment of new communication services on demand without the drawbacks of current AN proposals. These services are portable across heterogeneous hardware and software platforms. We have developed our proposal to proof-of-concept level. While our approach may be generalisable to permit dynamic deployment of active elements to clients and servers, we have chosen initially to focus on services that can be made transparent to existing servers and clients. This has many benefits, such as helping to solve extensive upgrade problems.

Our system provides a platform for flexible, value added service provision. However our work is incomplete. We have not yet developed dynamic binding of protocol elements. We need to develop and test a scalable system architecture. This will permit the dynamic discovery and location of DPSs. Likewise further work is required to address the manageability and robustness of the system, including specification of appropriate security mechanisms.

Section snippets

Overview

We describe an architecture which can support Application Level Active Networking. The system can be divided into a number of key components and concepts which are further described below. Fig. 1 shows how the components fit together. The key ideas are as follows.

In order to enhance the performance of an end to end communication proxylets are downloaded into Dynamic Proxy Servers which are strategically placed in the network. These proxylets are obtained from Proxylet Servers. We also intend

Implementation

The system that we have built thus far consists of a number of the components described above written in Java. However we feel that, while Java is useful for prototyping, the fundamental elements of our system are not dependent on Java. The following describes how the architectural components have been realised.

Control architecture

The central core of our system is the Dynamic Proxy Server which accepts code in the form of proxylets to execute on the users behalf. Typically one or more Dynamic Proxy Servers will aid in an end to end communication. Currently the locations of our DPSs are known. We therefore make explicit RMI connections to a particular DPS when we wish to start a proxylet. This explicit information is provided via the Control Interface (described below). However as large numbers of Dynamic Proxy Servers

Experiments

We have performed a number of experiments with our infrastructure using different applications or application services.

Performance measures

A potential issue with application level solutions is performance. We now present results that indicate the impact of our approach on bandwidth and latency.

Related work

Our work is significantly different to that embedded in existing WWW-based streaming tools. For example, the RealAudio tool from Progressive Networks supports streaming audio. Special purpose Audio servers and client players are used for the playout of the audio. Code cannot be pushed and pulled transparently, and support for adaptation is limited.

In the introduction we compared our approach to that of the Active Networks community. There is valuable work being carried out by that community,

Conclusions and future work

We have proposed a novel architecture and mechanisms for the achievement of Application Level Active Networks. We have developed and tested an initial implementation of some central aspects of the overall system, namely the Dynamic Proxy Server and downloading of proxylets. The feasibility of our proposals has been demonstrated via experimental implementation of a range of proxylets. These indicate that useful services can be downloaded dynamically. Our conclusion is that the proposed

Acknowledgements

This work is funded by British Telecom Laboratories. Our thanks go to Ian Marshall for the many fruitful discussions and suggestions. We thank Glen MacLarty for writing the webcache proxylet. We would also like to thank the anonymous reviewers for their many helpful suggestions.

Michael Fry is the Dean of the Faculty of Mathematical and Computing Sciences at the University of Technology, Sydney (UTS). He has a B.A. from Cambridge University, a M.Sc. from Imperial College London, and a Ph.D. from the University of Sydney. He has been with UTS for 18 years. Prior to that Michael worked in the IT industry developing real-time systems. Michael currently leads research projects investigating Quality of Service for multimedia delivery over the Internet, adaptive

References (12)

R. Barrett et al.
Intermediaries: new places for producing and manipulating web content
Computer Networks and ISDN Systems
(1998)
E. Amir, S. McCanne, R. Katz, An active service framework and its application to real-time multimedia transcoding, ACM...
S. Ardon, R. de Silva, R. Landfeldt, A. Seneviratne, Total management of transmissions for the end user: a framework...
T. Berners-Lee, The world-wide web initiative, Proc. Int. Networking Conf. (INET), San Francisco, CA, August...
S. Casner et al.
First IETF Internet audiocast
ACM Computer Communication Review
(1992)
A. Ghosh, M. Fry, Javaradio: an application level active network, 3rd Int. Workshop on High Performance Protocol...

There are more references available in the full text version of this article.

Cited by (70)

rfid in pervasive computing: State-of-the-art and outlook
2009, Pervasive and Mobile Computing
Citation Excerpt :
The principal enabler of this edgeware capability of the rfid stack is the event manager [20], which has the role to: In many ways the event manager is very similar in scope and function to peerage architectures encountered in overlay networks for example, peer-to-peer [21] and pervasive service delivery networks [22], and dynamic proxy servers in application-level active networks [23]. Seen in this way, rfid infrastructures consist of a network of event management peers, which are transparently accessible through a single interface, and coordinate their operation via super-peers at the network core.
rfid has already found its way into a variety of large scale applications and arguably it is already one of the most successful technologies in the history of computing. Beyond doubt, rfid is an effective automatic identification technology for a variety of objects including natural, manufactured and handmade artifacts; humans and other species; locations; and increasingly media content and mobile services. In this survey we consider developments towards establishing rfid as the cost-effective technical solution for the development of open, shared, universal pervasive computing infrastructures and look ahead to its future. In particular, we discuss the ingredients of current large scale applications; the role of network services to provide complete systems; privacy and security implications; and how rfid is helping prototype emerging pervasive computing applications. We conclude by identifying common trends in the new applications of rfid and ask questions related to sustainable universal deployment of this technology.
Towards a platform for wide-area overlay network deployment and management
2007, Computer Networks
Citation Excerpt :
However, we also imagine our system scaling to ubiquitous proportions, akin to current peer-to-peer networks (100,000 + nodes), which imposes a much tougher set of scalability requirements. We initially developed and demonstrated a number of proxylets that provide active services [1,2] including transcoders, protocol gateways and active caches. For example the IP telephony proxylet in Fig. 2 provides a gateway for voice calls between the internet and the telephony network.
This paper presents an experimental set of self-organising mechanisms to support application-level service overlays across the internet. These mechanisms include a base node discovery service and protocol that is resilient to node and network failure, which in turn is a basis for the construction of service-specific overlays. We have also developed a generic service overlay structured according to IP network latencies which uses the GNP virtual coordinate system. We discuss our results and experiences of implementing our platform using an active service infrastructure deployed on PlanetLab. While we have been successful in meeting many of our aims, we have identified some practical shortcomings with regard to latency estimation accuracy, and scalability of some of our mechanisms to internet size. The paper describes lessons learned from experimental development and deployment.
Performance-guided neural network for rapidly self-organising active network management
2004, Neurocomputing
Citation Excerpt :
Despite the difficulties of the approach [27], it has highlighted the important requirements for a feasible active network and encouraged other researchers to develop better alternatives to resolve them. The ALAN architecture was first proposed by Fry and Ghosh [15] to enable the user to supply JAVA-based active-service codes known as proxylets that run on an edge system (Execution Environment for Proxylets—EEPs) provided by the network operator. The purpose of the architecture is to enhance the communication between servers and clients using the EEPs that are located at optimal points of the end-to-end path between the server and the clients, without having to deal with the current system architecture and equipment.
We present a neural network for real-time learning and mapping of patterns using an external performance indicator. In a non-stationary environment where new patterns are introduced over time, the learning process utilises a novel snap-drift algorithm that performs fast, convergent, minimalist learning (snap) when the overall network performance is poor and slower, more cautious learning (drift) when the performance is good. Snap is based on a modified form of Adaptive Resonance Theory (CGIP 37(1987)54); and drift is based on Learning Vector Quantization (LVQ) (Proc. IJCNN 1(1990a)545). The two are combined within a semi-supervised learning system that shifts its learning style whenever it receives a significant change in performance feedback. The learning is capable of rapid re-learning and re-stabilisation, according to changes in external feedback or input patterns. We have incorporated this algorithm into the design of a modular neural network system, Performance-guided Adaptive Resonance Theory (P-ART) (Proc. IJCNN 2(2003)1412; Soft computing systems: Design, Management and application, IOS Press, Netherland, 2002; pp. 21–31). Simulation results show that the system discovers alternative solutions in response to significant changes in the input patterns and/or in the environment, which may require similar patterns to be treated differently over time. The simulations involve attempting to optimise the selection of network services in a non-stationary, real-time active computer network environment, in which the factors influencing the required selections are subject to change.
Enabling Software Defined Networks: Precursors, Evolution and Prognosis
2020, International Conference on Information Networking
A programmable structured peer-to-peer overlay
2009, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Cross-layer peer-topPeer traffic identification and adOptimization based on active networking
2009, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

View all citing articles on Scopus

Atanu Ghosh a B.Sc.(Eng) from Queen Mary College, University of London 1983 and a M.Sc. in Data Communication Networks and Distributed Systems from University College London 1991. He currently works in the School of Computing Sciences at the University of Technology, Sydney as a Research Scientist. He is working on a research project in collaboration with British Telecom Laboratories. The research is in the area of active networks. Previously he worked at University College London in the Department of Computer Science as Research Fellow. Originally on the MICE (Multimedia Integrated Conferencing for European Researchers) project which was a European project to promote multimedia conferencing over the Internet. The final project on which he worked was HIPPARCH (next generation HIgh Performance Protocol ARCHitecture). Before working at UCL he worked in industry on operating systems and networking.

View full text