Raschid Karabek

A transport protocol for native mode ATM networks design and implementation

Despite the increased bandwidth and the Quality of Service (QoS) capabilities of ATM networks, measurements show that only a fraction of the available bandwidth can be used by applications with current approaches for realizing data communications over ATM employing the TCP/IP protocol suite [1, 2]. This is largely due to the unsuited protocol mechanisms and the lack of QoS support on transport level. Native Mode ATM offers full access to ATMs Quality of Service capabilities in homogenous ATM networks and provides support for a wide range of information streams with different properties and requirements [6]. In order to exploit the capabilities of ATM networks, and to optimally use the available resources, in the network and end system, new transport protocols are inevitable. In this paper we propose the transport protocols ADTS (ATM Data Transfer Service), which uses efficient protocol mechanisms suitable for ATM systems with a minimum of protocol processing overhead. We discuss implementation issues and present performance results obtained with a reference implementation of the ADTS protocol.

A Software Architecture for Control and Management of Distributed ATM Switching

Distributed switching will become a key feature for flexible and inexpensive workgroup switches as well as for dedicated high performance networking solutions. The Philips Research Laboratories in Aachen have demonstrated the feasibility of distributed ATM switching by developing a ring-based ATM switch. After solving the hardware related problems we are now focusing our efforts on the design of a distributed software platform for switch control and management purposes.

Architectures for wireless ATM access

This paper presents a study on possible architecture design concepts for integrating wireless networks with ATM technology. We discuss four different solutions for wireless ATM access, starting with transparent cell-relaying over the air interface, thereafter focussing on ATM- and AAL-based interworking concepts, and finally discussing proxy approaches. Our analyses show that the performance benefits of higher layer interworking justify their comparatively high technical expenses.

Future Hypermedia Retrieval Systems and their Impact on Transfer Systems

Among the most popular distributed applications in todays networks are Hypermedia Retrieval Systems. Due to their rising popularity as well as performance limitations imposed by the underlying communication system the well known ‘World Wide Web’ is turning a ‘World Wide Wait’: Response times increase and at peak traffic times many sites are virtually unreachable. On the other hand emerging network technologies like ATM offer a guaranteed Quality of Service (QoS). Future Hypermedia Retrieval Systems will have to be able to reserve bandwidth and request certain timing constraints depending on document media type. Due to the properties of these new media, additional temporal requirements are not only imposed on the presentation (i.e. synchronisation of audio, video and textual information) but also on the transmission of these media streams. There is a desperate need for a uniform QoS architecture and interface for mapping presentation quality to communications requirements to efficiently support the requirements of future distributed multimedia systems and to make optimal use of the QoS capabilities of new networking technologies.

Open Switching for ATM Networks

This contribution discusses the way how Open Switching can be supported in ATM networks. First the elements of an ATM system that are relevant for call control and call handling are described. Then we evaluate which of the call control interfaces should be realized by which parts of the system, especially which signalling aspects are relevant for the software architecture on network side. Finally we propose a concept for the integration of external call control as part of the network control architecture. The paper aims at a switch control architecture that facilitates for providing advanced applications in ATM networks by external call control. The definition of standardized protocols and the design of Open Switching applications themselves are out of the scope of our work.

A native ATM protocol architecture design and performance evaluation

ATM networks offer the Quality of Service (QoS) guarantees required for future multimedia applications. Native mode ATM access offers full access to ATMs Quality of Service capabilities in homogeneous ATM networks and provides support for a wide range of information streams with different properties and requirements. The author presents a native ATM service architecture offering a common service access point for establishing virtual connections and additional transport layer functionality such as end-to-end error, flow and congestion control, datagram transfer and VC multiplexing. The author introduces the elements of the architecture, their functionality and capabilities. ATM networks are characterized by sudden changes in the available bandwidth due to the arrival or departure of constant bit rate (CBR) or variable bit rate (VBR) connections, and short term congestion in ATM switches caused by sporadic superimpositions of bursts from several sources. The author focuses on protocol mechanisms aimed at maximising effective end-to-end throughput (goodput) on data channels using ATMs best effort service, the unspecified bit rate (UBR) service.

Optimierung Von Transfersystemen

Die zunehmende Vielzahl neuer Applikationen motiviert aufgrund ihrer gehobenen Leistungsund Funktionalitatsanforderungen ein Redesign von Netzwerk-Protokollarchitekturen. Eine angeforderte Dienstgute (Quality of Service, QoS) soll nicht nur in lokalen homogenen Netzen, sondern auch uber Netzwerkgrenzen hinweg in einem globalen heterogenen Netz gewahrleistet werden. Eine Analyse der Syntax und Semantik existierender Transfersysteme (TCP/IP, OSI TP4/CLNP und XTP) hat ergeben, das diese nicht in der Lage sind, fur jede Anwendung die geeignete Protokollfunktionalitat zur Verfugung zu stellen. Es fehlen vor allem effiziente Protokollmechanismen zur Unterstutzung von Gruppenkommunikation und zur Vermeidung und zum Abbau von Uberlast. In diesem Artikel werden daher Ansatze zur Realisierung von Multi- casting basierend auf dem XTP-Bucket-Algorithmus sowie Erweiterungen und Untersuchungen eines auf Slow-Start/Tri-S basierenden Uberlastkontrollverfahrens vorgestellt.

Sprachübertragung über ISPNs. Messung und Modellierung

Zusatzlich zur klassischen Datenkommunikation mussen kunftige paketvermittelnde Netze (ISPNs, Integrated Services Packet Networks) in der Lage sein, Audio- und Videodatenstrome gemas ihrer harten zeitlichen Anforderungen zu ubertragen. Zur Entwicklung, Optimierung und Bewertung von ISPNs werden Simulationen eingesetzt. Ausschlaggebend fur die Qualitat der erzielten Ergebnisse sind realistische Lastgeneratoren fur Audio-, Video- sowie herkommliche Datenstrome. Bisherige Audio-Lastgeneratoren basieren auf Untersuchungen aus dem Bereich der Telekommunikation, die sich nicht auf ISPNs ubertragen lassen. Zur Modellierung von Sprachdatenstromen existieren zahlreiche Studien, die sich mit statistischen Analysen der Lan-gen von Sprach-(Talkspurts) und Stillephasen (Silence) beschaftigen. Herkommliche sprach-aktivitats-erkennende Systeme ermitteln diese Langen unter Verwendung relativ kurzer Analyseintervalle (5–10 ms). Die Talkspurt-und Silencelangen konnen dabei durch geometrische Verteilungen angenahert werden. Dieser Artikel beschaftigt sich mit der Modellierung von Sprachdatenstromen fur die veranderten Gegebenheiten paketvermittelnder Netze. Dazu wurde ein experimentelles Sprachaktivitats-Erkennungssystem, bestehend aus Digitalisierungshard-ware, Benutzeroberflache und interruptgesteuerter Analyse erstellt. Die Ergebnisse zeigen, das sich beim Ubergang auf Pakete als kleinste Analyseintervalle vollig andere Verteilungen ergeben. Hierzu bieten wir einen Erklarungsansatz anhand von Korrelationsbetrachtungen. Dieser deckt eine statistische Abhangigkeit zwischen Talkspurt-und Silence-Langen insbesondere bei kurzen Analyseintervallen auf. Desweiteren wurden die wahrend aktiver Sprachphasen auftretenden Stilleperioden zwischen Phrasen, Wortern bzw. Silben in die Modellierung integriert.