This paper considers the implementation of ABR functionality in a queuing point located in the access part of the network. The developed Explicit Rate flow control mechanism is the result of a balanced trade-off between implementation complexity and link load efficiency. Also, the ABR end-system behavior is implemented on a Network Interface Card. At the same time, a solution is presented to realize ABR in a service multiplexer built around a shared-medium bus. Experimental and simulation results are presented which show the main characteristics of ABR. It is demonstrated that a cost- effective implementation of ABR performs very efficiently.

HFC networks, today mostly used to broadcast traditional cable TV programs, will possibly play a major role in future broadband access network development. The structure of HFC networks is briefly described and bottlenecks in case of high load are identified. To improve performance and eliminate the bottleneck, the approach presented in this paper suggests the replacement of passive network elements like fiber nodes and signal amplifiers by ATM switches. This approach has been evaluated with respect to various network topologies and user scenarios. A simulation model used for the evaluation is presented within this paper and simulation results are discussed.

This paper presents an approach for carrying Internet transport PDUs directly over AAL5, discarding IP. The main advantage is an overhead reduction that can significantly impact upstream Internet traffic performance in asymmetric access network technologies such as HFC and ADSL. The TONA model proposed in the paper considers ATM to be at network layer-3, and tries to combine the best solutions given by previous models. Removal of the IP layer and the use of TCP over native ATM were achieved in a way that maximizes comparability with current standards and existent Internet applications, and allows the use of TCP by ATM applications.

Passive Optical Networks (PONs) are a promising future-proof solution for the broadband access network. A topic to be considered in planning a PON architecture is the position of the branching devices (splitters). Some technical and economical considerations about the splitter location are given in order to evaluate the advantages and the critical issues of the different possibilities.

This paper reports on the control systems for upstream optical repeaters in a long range high-split ratio SuperPON access network. Optical amplification means of fast switching semiconductor optical amplifiers is used. Two methods to control the drive current of the optical amplifiers are described: a feed forward self-triggered scheme and a topology using a priori known talker identity information. Trade-offs between both systems will be discussed. The first alternative was used in a lab demonstrator set-up, while the second will be used in a small field trial.

We propose a new receiver structure for optical code-division multiple-access (CDMA) networks avoiding data collisions. These collisions result from the used addressing scheme, and they occur, if more than one transmitter sends its data packets to the same receiver. By using a receiver for multiple data packets and considering the capture effect we show, that the effect of data collisions can be reduced significantly. Moreover, we analyze the performance of this scheme using packet success probability and network throughput.

Hybrid analog/digital Sub-Carrier-Multiplexing (SCM) technique is effective for broadband optical access networks designed for video broadcasting and interactive services. These systems are flexible in service management and channeling can be optimized to exploit the available bandwidth. A significant impairment in these systems is introduced by optical fiber multiple reflections, which give rise to an interferometric noise affecting the performance. So far, this effect has been evaluated assuming that the delay between transmitted and reflected optical signals is so large that such signals can be considered uncorrelated. In this paper, the interferometric noise due to optical fiber reflections is analyzed in SCM hybrid analog/digital systems independently of the position of the reflection points. The proposed model leads to a closed expression for the power spectral density of such noise source, which provides useful information for the system design and permits an accurate performance evaluation.

The presentation describes a new scenario for how to transform the existing telephone access network into a real broadband network. The objective is to introduce a low-cost network evolution scenario, which makes maximal use of the huge investments connected to this kind of networks. The existing network is sub-divided into a small cell network, each covering an area with an approximate radius of 1 kilometer. Inverse multiplexing of xDSL systems is used to aggregate bandwidth in the range of 100 to 1000 Mb/s between nodes in the network. VDSL modems able to carry about 25 - 50 Mb/s, are used on the last section from the node to the customer premises. These modems provide all households within each geographical cell with real broadband access to a large common bandwidth.

DSL technology enables very high bandwidth transmission in a point-to-point fashion from a customer's premises to a central office (CO), wiring center, or other logical point of traffic aggregation. Unlike many technologies that enable broadband Internet access, DSL technology does not determine a specific architecture to be deployed at either the customer's premises or in the service/access provider's network. In fact, DSL technology can be used in conjunction with a variety of network architectures. While being agnostic regarding to higher-layer protocols, there are still several critical 'protocol specific' issues that need to be addressed when deploying DSL as a solution for IP (Internet/intrAnet) access. This paper will address these issues and present a range of network architectures that incorporate DSL technology. This paper will only focus on those architectures that enable IP access. These architectures are divided into three categories: Traditional Dialled Model (TDM), frame-based (Frame Relay/Ethernet), and cell-based (ATM).

As interactive services require more bandwidth, high speed access networks must follow this trend. Although a Hybrid Fiber Coax (HFC) system provides a large bandwidth, many problems occur due to the accumulation of the ingress and impulse noise. In this paper the properties of the ingress and impulse noise will be explained based on a measurement campaign on different networks. The most striking result of the measurement campaign is the large dynamic and time- dependent behavior of the ingress noise. This dynamic behavior is an important factor in designing an upstream communication system. We show that for this reason optimized receivers will contain adaptive equalizers. This is illustrated by the performance analysis of a TDMA and an OFDM-CDMA system. Both systems are compared under the same environment. The results of this comparison shows how both can obtain the same performance. Considering the implications of both the TDMA and the CDMA system when using an equalizer in the head-end receiver, requirements for a total HFC network architecture are given. Especially synchronization and initialization issues must be addressed.

This paper introduces the ETSI Project on Broadband Radio Access Networks (EP-BRAN). BRAN systems will be used for local area applications with limited mobility (HIPERLAN/2); fixed access with area coverage in urban and rural areas (HIPERACCESS); and short range high-speed point-to-point links (HIPERLINK). They will support transport of either IP or ATM protocols, supporting managed quality of service. Such systems are needed to provide access to the future broadband core networks supporting multi-media applications. The paper addresses the motivation and market demand for broadband radio access networks, the objectives and scope of the Project, the operational and technical requirements, the types of networks to be standardized, the scope of the standards, the issue of spectrum and the Project schedule.

Wireless ATM provides a framework for next-generation wireless communication networks supporting QoS based multimedia services. A proof-of-concept prototype of a Wireless ATM network, termed WATMnet, has been developed at NEC USA C&C Research Labs in Princeton, NJ. In this paper, we discuss the implementation of radio access protocol layers in WATMnet. We also present results from an experimental evaluation of the various protocol layers and their effect on higher layer performance.

As telecom providers introduce new and more sophisticated services the necessity of a global, unified view of the network infrastructure becomes demanding. Today, heterogenous backbone networks are interconnected in order to provide global connectivity. Due to technological impairments the cost of network operation, the maintenance complexity and the overuse of resources are extremely high under the goal of supporting the diverting customer requirements. We propose a scheme for ATM QoS support in such heterogenous, multi-domain, multi-technology network environment. The objective is to optimize users' and networks' profits by giving them the opportunity to satisfy their requirements. Our approach introduces a manager able to take routing decisions supporting quality of service guarantees for the customers, while making efficient use of network resources.

This document presents a MIB (Management Information Base) for use in the end-to-end management of networks which utilize ADSL (Asymmetric Dual Subscriber Line) technology for the 'last mile' (i.e. communication between the PSTN central office and users' premises.) The 'Container' MIB is useful to the Network Management System (NMS) in abstracting information derived from lower network layers and in cross-referencing elements of disparate network layers. The Container MIB is described here in the context of networks which use ATM (Asynchronous Transfer Mode) technology for the backbone. Conceptually, the Container MIB is above the ATM, ADSL, and entity MIBs in the management hierarchy, and bridges between ATM and ADSL network segments. We demonstrate how the Container MIB can be used in the implementation of several unique network management functions endemic to ATM/ADSL networks.

In Some kind of networks like PON (Passive Optical Networks) and HFC (Hybrid Fiber Coaxial) networks, the information from the central station (headend) to the terminals is broadcast to all of them. So each one of them besides its own data gets also the data destined for the others. Under these conditions privacy and confidentiality do not exist. The usual way to overcome this drawback is conventional encryption of the data between headend and terminals. This method besides making the communication procedure more complex, has also the drawback that traffic data cannot be protected, since the destination field of each cell cannot be encrypted. Here another method is presented which, although not encryption in the conventional meaning of the term, secures a high degree of individuality and confidentiality. The basic idea is to use different multiplexing patterns for each connection so that each terminal can demultiplex only its own data. The method is suited for the high data speed of the PON networks and maybe preferable to conventional encryption.

This paper describes a new scheme for dispersion compensation called nonlinearity tolerant dispersion compensation scheme (NTDCS). The scheme is characterized by a short piece of dispersion compensating fiber (DCF) in front of the transmission fiber and a second DCF after it compensating for the rest of SSMF dispersion. The NTDCS shows much better eye restoration than precompensation or postcompensation. It is characterized by a shift of a postcompensated dispersion map towards smaller values. A shift of -280 ps/nm is found to be best choice. The tolerance against SPM is demonstrated by simulations with varying channel powers. Up to 10.5 dBm amplifier output power can easily be tolerated for a 8 X 80 km transmission.

Advanced manufacture of DWDM multidielectric narrow-band optical filters designed for broadband technologies are presented and discussed. Three-cavity structures of 2 nm half- maximum bandwidth pass-band filters in the 1550 nm range with more than 94% at peak transmission and 4 nm width at 20 dB of attenuation is presented. It has been designed for 400 GHz channel spacing.

In the near future, satellite networks will be an essential part of the upcoming Global Information Infrastructure (GII). These satellites will be interconnected by high bandwidth inter-satellite-links using lasercomm technology. Contraves Space has designed a family of lasercomm terminals for different satellite network topologies but based on common technology key elements: (1) Coherent Syncbit data transmission scheme for highest receiver sensitivity and, hence, best system efficiency. (2) Laser diode pumped solid state lasers for best reliability figures. (3) A combined coherent receiver-frontend and coherent tracking sensor for sun outage-free operation. As a first terminal the SROIL (Short-Range-Optical Inter-Satellite-Link-Terminal) was completed in 1997 and undergoes currently an extensive phase. Two SROIL terminals will be embarked in 1999 on two LEO spacecrafts and perform a full high data rate transmission demonstration in-orbit.

An efficient method for evaluating the error probability of optical ASK/DD systems subject to interferometric noise is presented. The receiver decision variable is statistically described by its moment generating function (mgf). The theoretical results, obtained with the aid of the new derived mgf, are in good agreement with experiment, employing directly modulated light sources, while the common used Gaussian statistics for the photocurrent yields larger power penalties. The analysis takes into consideration polarization statistics, photodetector shot noise, non-ideal extinction ratio, and receiver thermal noise. Error probabilities are calculated using the saddlepoint approximation which is numerically simple and gives accurate results.

A monolithic Clock and Data Recovery (CDR) circuit for SDH STM-16 (2.5 Gb/s) digital receivers has been designed and fabricated using Maxim GST-2 27 GHz Silicon bipolar technology. The main functions carried out by the IC are: signal amplification (40 dB) and limitation, clock recovery and decision. The design is intended to achieve a complete 2.5 Gb/s receiver by using the IC and a low noise preamplifier (transimpedance stage), mounted in a DIL package. The integrated circuit comprises about 400 active devices, used both for analog and digital blocks, and uses two supply voltages of 5 and -4.5 V. The input port is decoupled by external capacitors and matched to 50 (Omega) using on-chip resistors, whereas clock and data outputs are open collector type. The die size is 2 X 2 mm² and the chip has been packaged using a TQFP 48 pins plastic package. Measurements under 2³¹-1 PRBS data stream have shown an input sensitivity below 5 mVpp, rms output jitter below 7 ps and total power consumption of 0.8 W.

In this paper a network architecture for all optical packet switching is presented, suitable for both long distance and metropolitan application. Overall network architecture, protocol layering, functional sub-blocks required and wavelength management are discussed with reference to the latest results obtained within the ACTS Project KEOPS (Keys in Optical Packet Switching).

The influence of in-band crosstalk on the error performance of all optical networks with different topologies is studied. A statistical crosstalk model is used for evaluating the bit- error rate. The model accounts for optical preamplification. We present a network topology having the best performance while using the largest transmission path.

All-optical Wavelength Division Multiplexing (WDM) networks are believed to be a fundamental component in future high speed backbones. However, while wavelength routing made circuit switching in WDM feasible the reality of extant optical technology does not yet provide the necessary devices to achieve individual optical packet switching. This paper proposes to achieve all-optical packet switching in WDM Wide Area Networks (WANs) via a novel technique, called slot routing. Using slot routing, entire slots, each carrying multiple packets on distinct wavelengths, are switched transparently and individually. As a result packets can be optically transmitted and switched in the network using available fast and wavelength non-sensitive devices. The proposed routing technique leads to an optical packet switching solution, that is simple, practical, and unique as it makes it possible to build a WDM all-optical WAN with optical devices based on proven technologies.

In this paper, we consider the optimal design of all-optical networks based on wavelength-division multiplexing (WDM). Given a static channel demand, we present a new method to minimize the numbers of optical fibers on each link which are at least necessary to transmit the demanded traffic via the network. In order to minimize the necessary number of fibers, the well known problem of routing and wavelength assignment (RWA) has to be solved, too. Therefore, we present a new approach to the solution of the RWA problem, which bases on the combination of a genetic algorithm and a simple heuristic. The efficiency of our new method is finally demonstrated by the application to a European backbone network.

In this paper, the restoration strategies with QoS-guarantee for optical network services are introduced. For that purpose optical network services in WDM networks are classified according to the quality degree and restorability required. In particular, the self-healing aspects of service restoration are studied. The simulation results for networks with arbitrary topology and service quality degree as well as for service-specific restoration methods are presented and studied on their applicability.

Optical wavelength division multiplex systems are being installed today. Optical networks will be implemented in the near future. However, little is known about the operational aspects and requirements. In the projects PHOTON and MOON not only an optical network with more than 500 km transmission length is established but also processes are being developed that might be used for installation and operation of future networks. The presentation gives an overview about the projects, focuses on the field trials and the envisaged tests and applications.

Recent progress of optical technology will soon lead to a variety of innovative products available on the market for designing optical networks. These products will make it possible to realize network layer functionality, such as transmission, multiplexing, routing and performance assessment directly in the optical layer. One of the reasons behind this evolution is the potential cost reduction of the overall system obtained by introducing an optical layer, over the extant transmission network, supporting the growing demand for broadband services. The paper analyses the cost reduction achievable by this emerging technology in a regional network with about two million users. The analysis is carried out taking into account a number of relevant factors, including the total fiber length, the number of wavelengths per fiber, the transmission rates, the line terminals, the structure of the nodes, considering both the optical and the electrical components, that are necessary to support the traffic scenarios expected over the next five years and beyond.

In order to improve the overall availability of SDH-based ATM transport networks in a cost-effective way, a survivability strategy coordinating the recovery schemes at both ATM and SDH layers has to be devised. This paper proposes a new survivability concept, common pool survivability, for multi- layer ATM over SDH transport networks. The basic idea is to share a common pool of spare resources amongst the ATM and SDH recovery schemes. This can be accomplished by supporting the ATM spare link capacity through unprotected preemptible connections in the SDH network. It is shown on a set of sample networks that common pool survivability yields considerable cost savings when compared to traditional multilayer survivability designs.

The protection aspects represent a fundamental point in designing a modern transport network. For this reason, SDH technology was developed taking into maximum account the survivability requirements. The evolution of backbone transport network towards the WDM technology will have also to take into account these aspects. In particular this paper highlights the open issues to coordinate the interworking between the SDH and WDM network layer, the trade-off between transparency and manageability, and presents a possible evolution scenario of how WDM technology could integrate into SDH-based transport networks. The increasing installation of WDM systems from point to point system to multi wavelength optical ring and even to optical cross connect will force the definition of new planning and protection strategies.

With the scale of disk array, more and more disk arrays are becoming stand alone. There are many advantages of using stand alone disk array, such as easy to install, upgrade and maintain, scalable. But the communication between host (called master) and disk array controller (called slave) is the main bottleneck problem in the system. The requirements for communication between master and slave are much higher. The bandwidth needed is well beyond that interface can supported and new ideas and methods are needed. In the communication between master and slave, higher performance is required for optimal utilization of the bandwidth. One of the promising technology to meet these requirements is to use the high speed data channel between master and slave. In this paper, a new method of using high speed data channel in the system of disk array is proposed. Some design issues of disk array using high speed data channel are also analyzed.

We have developed a set-top box with video encoder (STB-VE), in order to provide STB users with bi-directional multimedia applications such as video conferencing, in addition to traditional STB applications such as Video On Demand (VOD). This paper comprises of two parts. The first part describes the design of the STB-VE hardware and software. The second part describes the application of STB-VE to videoconferencing. We have designed a new protocol called the Multimedia Session Control Protocol (MSCP) for conversational multimedia applications such as video conferencing. We have developed the video conferencing system using MPEG1, by implementing MSCP on STB-VE.

The possibilities for video transmission via the Internet have improved a great deal over the last 2 years. The state of the art in this field is Streaming Video which allows video data to be transmitted via the network in real time and simultaneously displayed. Streaming Video technology is being developed continuously. Our examinations of this technology have shown that special protocols are preferable for real time applications. Technologies such as Resource Reservation Protocol (RSVP) and Asynchronous Transfer Mode (ATM) in particular can make a significant contribution to the Quality of Service provided during video transmission by enabling bandwidth reservation. This paper analyzes specific requirements for video-on-demand and those for live transmission. Based on the resulting ideal properties, we propose a generic architecture for a live video server which can be flexibly adapted to various transfer protocols.

The paper presents new method for compression of correlated image sets. It is known that Karhunen-Loeve (KL) transform is most optimal representation for such a purpose. In our paper we use recently suggested fast algorithm of KL basis construction for compression of correlated image ensembles. The approach is based on fact that every KL basis function give maximum possible contribution in every image and this contribution decreases most quickly among all possible bases. So, we lossy compress every KL basis function by lossy DCT coding with essentially different loss that depends on the functions' contribution in the images.

Once upon a time, or more precisely during February 1988 at the CCITT Seoul plenary, and definitely arriving as a revolution, ATM hit the hard-core B-ISDN circuit-switching gang. Initiated by the Telecoms' camp, but, surprisingly, soon to be pushed by computer minded people, ATM's generic technological history is somewhat richer than single-sided stories. Here are two classical elements of that history: Firstly, together with X.25, ATM suffers from the connection versus datagram dichotomy, well known for more than twenty years. Secondly, and lesser known, ATM's use of cells in support of the 'I' of B-ISDN was questioned from the very beginning by the packet switching camp. Furthermore, in this context, there are two other essential elements to be considered: Firstly, the exponential growth of the Internet and later intranets, using Internet technology, sparked by the success of the Web and the WINTEL alliance, resulted in a corresponding demand for both aggregate and end-system network bandwidth. Secondly, servers, historically restricted to the exclusive club of HIPPI-equipped supercomputers, suddenly become ordinary high-end PCs with 64-bit wide PCI busses -- definitely aiming at the Gigabit. Here, if your aim is for Gigabit ATM with 5000-transactions per second classical supercomputers, a 65K ATM MTU -- as implemented by Cray -- might be okay. Following Clark and others, another part of the story is the adoption and redefinition, by the IETF, of the Telecoms' notion of 'Integrated Services' and QoS mechanisms. The quest for low-delay IP packet forwarding, perhaps possible over ATM cut-throughs, has resulted in the switching versus/or integrated-with-routing movement. However, a blow for ATM may be the recent results concerning fast routing table lookup algorithms. This, by making Gigabit routing possible using ordinary Pentium processors may eventually render the much prophesized ATM switching performance unnecessary. Recently, with the rise of Gigabit Ethernet, many of the elements mentioned above are now being presented by standard 'Gigabit Ethernet and Gigabit ATM -- friends or foes' conferences. In- depth analyses are given concerning the canonical elements of such a setting: legacy, new use requirements, manageability, security LAN-WAN, architectures, standards, technologies and products, complexity, evolution-transition strategies, manufacturers, player organizations etc. Often in such conferences, Fiber Channel, being one of Gigabit Ethernet's physical media, is presented as the only other Gigabit LAN technology. In an attempt to sum up: Given the present state of ATM deployment measured in terms of functionalities and sophistication, after ten years of CCCITT/ITU and almost as many years of ATM Forum effort, does the question still being asked now represent the answer -- ATM is or was a mistake there some elements still missing? Here's a technical and a political example:

A new optical local area network that combines dense Wavelength Division Multiplexing (WDM) techniques and coherent transmission benefits is presented. A new multichannel protocol is specified based on Ethernet, using WDM to carry multiple channels on a single fiber. The protocol is simulated to evaluate the performance under congestion conditions and to compare it with other protocols having the same total capacity. Some suggestions are made for congestion control in high-capacity networks and different channel-jumping strategies are proposed. A network architecture is proposed with tunable transmitters and multichannel receivers, and with a common local oscillator.

This paper describes and analyzes SR³ (Synchronous Round Robin with Reservations), a collision-free medium access control protocol for all-optical slotted packet networks based on WDM multi-channel ring topologies where nodes are equipped with one fixed-wavelength receiver and one wavelength-tunable transmitter. SR³ achieves an efficient exploitation of the available bandwidth, guarantees a throughput-fair access to each node, and allows nodes to reserve slots, thereby permitting a tight control on access delays. The latter feature makes the proposed protocol well suited to provide support to traffic classes with different QoS requirements. Simulation results show that SR³ provides very good performance to guaranteed quality traffic, bringing significant performance improvements also to best-effort traffic.

Switchlets result from partitioning the resources, both physical and logical, of a switch. Each partition is a switchlet. Switchlets from different physical switches may be interconnected together in a manner consistent with the physical connectivity of the real switches. Each set of interconnected switchlets can be controlled by its own system; this control system is called a control architecture. Such an arrangement results in a set of virtual networks, each with its own control architecture. These control architectures may be identical, that is they may be different instances of the same control architecture, or they may be radically different. In this paper we examine how switchlets can be used to provide functionality which is either impossible or problematic to provide using conventional approaches to network control. While the examples being used will be based on ATM, it should be clear that application to any connection oriented network is straightforward. The switchlet technique is protected by patent.

Virtual LANs (VLANs) allow to interconnect users over campus or wide area networks and gives the users the impression as they would be connected to the same local area network (LAN). The implementation of VLANs is based on ATM Forum's LAN Emulation and LAN/ATM switches providing interconnection of emulated LANs over ATM and the LAN ports to which the user's end systems are attached to. The paper discusses possible implementation architectures and describes advanced features such as ATM short-cuts, QoS, and redundancy concepts.

In previous works, we have introduced the notion of supranet. Supranets are virtual networks -- private to a group of users -- that can be built on top of a physical network (e.g., the Internet) by any user of such a network making use of an appropriate software toolkit. We expect supranets to be used in the future by many different types of groups, each with different requirements in terms of security. Since it is necessary to meet very different needs, one of our design goals has been to allow the users to choose the security level they feel is more appropriate in each case. This can be done, for instance, by specifying appropriate options at supranet construction time. This paper discuses the ideals on which supranets are based as well as the security issues to be considered when designing supranets on top of the Internet.

The term fully integrated systems refers to the interconnection of all the Information Technologies and Telecommunications in distributed environments. It is a rapidly growing interest in finding and standardizing a communication protocol that could merge the vast amount of services existing today. This study addresses the advantages in using ATM as a backbone in a real time distributed environments.

Field buses have been developed to fulfill real-time requirements of applications related to factory automation. However, they are limited in their extension, and can only cover geographical areas such as factory halls or buildings. In order to reduce this drawback a possible solution is to connect field bus segments using interworking devices and a backbone network. This backbone could be an ATM network. Within this paper we concentrate on the interworking of PROFIBUS segments. The main topic is to investigate performance aspects of such an interworking concept. From PROFIBUS point of view the impact on the parameterization of the PROFIBUS system is of special interest. Simulations of an example PROFIBUS system were performed in order to cover this aspect, to evaluate the feasibility of the presented approach and to show its limitations concerning the range of PROFIBUS applications it is suited for. Another main topic in this concern is the choice of appropriate QoS parameters for the ATM connections.

This paper explores how dynamic quality-of-service may be supported through RSVP over IPSOFACTO within a network of ATM switches. IPSOFACTO is a methodology fur supporting IP directly over ATM switches. Application endpoints communicate exclusively via IP and no ATM signalling is involved. The application's dynamic QoS requirements are communicated by a user through a novel user interface, called CockpitView, and these are signalled through the network via RSVP running directly on ATM switches. Models to accommodate receiver heterogeneity in this environment are described. A Traffic Control module interfacing to both RSVP signalling and IPSOFACTO is presented. Finally, an experimental setup over the MAY/BALI networks integrating IPSOFACTO, RSVP and CockpitView is presented.

Today we are faced with an increasing variety of networks and end systems. The resulting heterogenous environment imposes new challenges on communication support for multimedia and collaborative applications. Heterogenous group communication is a key issue for the proper support of such applications. The approach outlined in this paper forms a first step towards an active multicasting network (AMnet) that is capable to provide flexible, scalable and efficient support for heterogenous group communication. RSVP is used as signaling protocol and QoS filters form the active components in the networks nodes. They scale the QoS down to the individual requirements of subgroups or single members within a heterogenous group communication. Therefore, they remove information (e.g., color information) from continuous media streams, such as video streams. The basic architecture and concept as well as a prototype implementation and some measurements are presented in this paper.

In order to solve the current problems for QoS controllable multimedia communication applications, this paper analyzes generic requirements for multimedia communications, clarifies basic roles and functions and the efficient assignment of them, and proposes a new architecture. This paper takes pragmatic approaches and focuses on the mechanism on how to provide QoS control features for multimedia communication applications with a minimum of change in current traditional programming styles. These proposed ideas and architecture have been implemented into the prototype system, on WindowsNT with real-time MPEG encoding and decoding multimedia data communications. This paper describes this implementation and the result of the architecture and ideas.

Many video and audio sessions are transmitted via the Multicast Backbone (MBone) medium day by day. The Integrated Services architecture of the Internet will make it now possible to transmit and receive these sessions in a 'guaranteed' quality of service. To allow this in heterogeneous networks, consisting of, e.g. Ethernet and ATM, an interaction between the QoS architectures of the Internet and ATM is necessary. However, they are very different, so mapping them is a difficult task. In this paper, the implementation of an interaction approach for the QoS architectures developed for the Internet and for ATM, and the extension of the MBone VCR for recording and playing back MBone sessions in a higher quality, is described. This system allows the QoS supported recording and off-line playback of MBone session in heterogeneous IP/ATM networks.

This paper presents an algorithm to control QoS for distributed multimedia applications. The algorithm approaches the problem in an integrated way solving both the problems of the network bandwidth and the load at the workstations. It interacts with the applications by an abstract sequence to describe values for QoS that are more meaningful to application programmers than the low-level entities such as cell loss rate, throughput, etc. Some experimental results of the algorithm are presented.

Multimedia applications require besides basic communication services the particular support of guaranteed end-to-end services. In addition, these services need to be of a multipoint nature, where group communications take place in a guaranteed fashion as well. Therefore, a Quality-of-Service- driven specification of application requirements for multipoint scenarios and the guaranteed provision has been developed. At the same time, multipoint dynamics in terms of joining or leaving participants have been integrated, still guaranteeing end-to-end performance. In contrast to other communication frameworks, Quality-of-Service mechanisms and multipoint communication support are not considered as orthogonal, but an integration of these aspects leads to an efficient and highly flexible approach. Resource reservation, flexible protocol stacks, and heterogeneous network support are provided to guarantee multipoint multimedia application communication services in an end-to-end fashion.

Recent research efforts have concentrated on determining how the distributed workplace can be transformed into a shared virtual environment. Interaction among people and process virtual worlds has to be provided and improved. To enhance the usability of our virtual collaborative environment we integrated a multicast communication environment. With the availability of global information highways, 3D graphical intercontinental collaboration will become a part of our daily work routine. This paper describes the basics of our network infrastructure and the multicast support. As a proof of concept, a virtual world scenario is also presented in this paper.

The Web has rapidly grown to become the standard for documents interchange on the Internet. At the same time the interest on transmitting continuous media flows on the Internet, and its associated applications like multimedia on demand, is also growing. Integrating both kinds of systems should allow building real hypermedia systems where all media objects can be linked from any other, taking into account temporal and spatial synchronization. A way to achieve this integration is using the Corba architecture. This is a standard for open distributed systems. There are also recent efforts to integrate Web and Corba systems. We use this architecture to build a service for distribution of data flows endowed with timing restrictions. We use to integrate it with the Web, by one side Java applets that can use the Corba architecture and are embedded on HTML pages. On the other side, we also benefit from the efforts to integrate Corba and the Web.

We introduce an original scheme that turns the World Wide Web (WWW) into a social place. Our vicinity-based approach is to show the Web's user space to the people surfing on the WWW. Being aware of other Web surfers, people browsing the Web may invoke synchronous communication by a single mouse-click -- surfing the Web turns from a lonely affair into a joint experience. We present the basic theoretical concepts of our approach and give a short overview over CoBrow, a collaborative browsing service that implements the concepts presented in this paper. We also try to assess the social impact of services like CoBrow, which bring people together in a virtual world.

Routers are key building blocks in networks. They need to cope with high data rates in the range of multiple gigabit per second that are flowing through them. Therefore, specifically performance critical functions should be implemented in dedicated hardware units in order to speed up the forwarding task. These units can be embedded within regular workstations or into dedicated router architectures. This paper we address one of the most performance critical components of a router, the routing table and its access and search mechanisms. Earlier work has shown that this is more critical with respect to the resulting performance than IP processing itself. A simple but efficient organization of the routing table using binary trees and off-the-shelf SRAMs is presented together with a suited search algorithm.

The performances of some popular routing algorithms used for multicast communications are extensively simulated and analyzed in comparative terms, assuming the Bidirectional Manhattan Street Network as a valued test-bed. Multicast traffic is inserted over a substrate of point-to-point communications, contributing to increase the value of some cost parameters, suitably defined to express the routing efficiency. Both the cases of uncorrelated and bursty traffic are considered. The analysis confirms that the superiority of an algorithm over the others cannot be claimed since it depends on the input variables and on the criterion used for performance evaluation. Anyway conclusions can be drawn under specific assumptions, and they are usable in the framework of the network design.

Multimedia electronic mail and active electronic mail are very well known concepts. Multimedia electronic mail is concerned with the transmission of multimedia information like images, audio and video objects within email. Moreover, active email (or enhanced email) incorporates programs in ordinary electronic mail. The active email message automatically executes at the email receiver's site and, for instance, orchestrates the presentation of the multimedia information. This kind of message can be regarded as a subtype of a mobile agent which as a user interface and which gets into contact with the enduser. Based on a realized prototype ActiveM3, the related concepts for multimedia authoring and synchronization within the active email approach are introduced in this article.

Ism is a new type of authoring environment, based on a digital library of consistent data structures, around which associated tools are grouped to support the creation, manipulation, and archiving/retrieval for the different syntactic, semantic, and semeiotic level of video material in the distinct phases of the video production process. A4SM is supposed to be used by all experts involved in realizing a film. The aim is to seamless integrate the system into the video production process. A4SM improves the quality of the production process itself -- because of the integrated management of all involved elements -- but also increases the exploitation of the gathered material.

This paper presents an approach for authoring training applications. In this approach we think an author can undertake two roles: the meta-authoring role and the authoring role. A meta-author selects educational material and proposes lessons to be used later by an author in order to build specific educational applications. These two roles are supported by the meta-authoring tool and the authoring tool, which both form the authoring environment. The system also comprises a repository with generic and re-usable data, and a delivery environment. Resulted applications consist of flexible educational sessions, which are adjustable according to the learner's profile and overcome several problems, such as disorientation, usually encountered in hypermedia applications. Thus, authors can easily and effectively build educational applications, trainees' progress is automatically assessed, and reusability is achieved.

The paper presents new method for compression of correlated image sets. It is known that Karhunen-Loeve (KL) transform is most optimal representation for such a purpose. In our paper we use recently suggested fast algorithm of KL basis construction for compression of correlated image ensembles. The approach is based on fact that every KL basis function give maximum possible contribution in every image and this contribution decreases most quickly among all possible bases. So, we lossy compress every KL basis function by lossy DCT coding with essentially different loss that depends on the functions' contribution in the images.

The use of software codecs for video compression becomes commonplace in several videoconferencing applications. In order to reduce conflicts with other applications used at the same time, mechanisms for resource reservation on endsystems need to determine an upper bound for computing time used by the codec. This leads to the demand for predictable execution times of compression/decompression. Since compression schemes as H.261 inherently depend on the motion contained in the video, an adaptive admission control is required. This paper presents a data driven approach based on dynamical reduction of the number of processed macroblocks in peak situations. Besides the absolute speed is a point of interest. The question, whether and how software compression of high quality video is feasible on today's desktop computers, is examined.

In this paper, we present the design of QoSMail, QoS email application demonstrating a flexible model for using simple declarative requirement-statements and awareness of system properties to dynamically select policies governing message delivery. We believe that these ideas may apply to other cases as well such as binding of stream interfaces. Foundations of the QoSmail design are a framework for describing Quality of Service requirements, a framework for describing the relevant attributes of the environment, i.e. the resources available for transport, and a framework for policy selection, based on QoS-requirements and environment attributes.

In this paper we will introduce a method of analyzing images, a criterium to differentiate between images, a compression method of medical images in digital form based on the classification of the image bit plane and finally an algorithm for adaptive image compression. The analysis of the image content is based on a valuation of the relative number and absolute values of the wavelet coefficients. A comparison between the original image and the decoded image will be done by a difference criteria calculated by the wavelet coefficients of the original image and the decoded image of the first and second iteration step of the wavelet transformation. This adaptive image compression algorithm is based on a classification of digital images into three classes and followed by the compression of the image by a suitable compression algorithm. Furthermore we will show that applying these classification rules on DICOM-images is a very effective method to do adaptive compression. The image classification algorithm and the image compression algorithms have been implemented in JAVA.

In this paper, we propose framework for dynamically reconfigurable video codec. We introduce the concept of a virtual codec and a virtual tool to facilitate the use of multiple codec structures and multiple coding tools in a single video codec. Existing coding standards as well as new codec structures and coding tools can be integrated seamlessly in the proposed codec. The experimental results show that when implementing a standard codec such as H.263 the proposed codec can achieve a comparable performance to dedicated standard codecs. The results also demonstrate the codec's ability to use multiple coding tools and codec structures which are useful for implementing MSDL for MPEG-4 video.

Three recently introduced technologies: a new Autosophy information theory; fast Content Addressable Memories; and Packet Switching protocols, are combined in data compression chipsets to improve network throughput with real-time lossless text and image compression. Compared with the cost of putting new cables into the ground or launching new satellites, data compression chipsets offer a much less expensive alternative for improving the bandwidth of broadband communication networks. Data compression is based on a new Autosophy information theory in which, in contrast to Shannon's theory, communication is determined only by the data content. In addition to high lossless data compression, this also provides virtually unbreakable 'codebook' encryption and easier communication via packet switching networks. The Autosophy theories provide eight known classes of self-learning Omni Dimensional Networks. Only the serial network is explained here for use in compressed text and video communications. A new Content Addressable Read Only Memory (CAROM) may increase data compression speed to more than 20 Million symbols per second, fast enough for virtually any network speed.

We have been designing and implementing an object-relational multimedia database. A new data model was first developed, which represents various types of media entities, and temporal/logical relationships among these entities. Thereafter, a three-layer object-relational database infrastructure was proposed in our project, in order to support the proposed data model. In this paper, we concentrate on the search and query mechanisms designated for our database infrastructure. In particular, we discuss a new query language, which is developed as a multimedia extension of SQL3. This query language provides a greater expressive power, not only on simple/complex media entities, but also on various relationships among different entities. Consequently, it supports both content-based and structure-based queries. We are currently implementing the query language in our object- relational multimedia database prototype system based on the Illustrar. The fundamental structure of our query language, as well as main features of Object Manager and Query Manager of our prototype system, are also presented in the paper.

This paper describes the first stages of a research project that is currently being developed in the Image Processing Group of the UPM. The aim of this effort is to add video capabilities to the Storage and Retrieval Information System already working at our premises. Here we will focus on the early design steps of a Video Information System. For this purpose, we present a review of most of the reported techniques for video temporal segmentation and semantic segmentation, previous steps to afford the content extraction task, and we discuss them to select the more suitable ones. We then outline a block design of a temporal segmentation module, and present guidelines to the design of the semantic segmentation one. All these operations trend to facilitate automation in the extraction of low level features and semantic features that will finally take part of the video descriptors.