Introduction "Multimedia synchronization" is needed to ensure a temporal ordering of events in multimedi a systems . There are two parts to this problem : the definition of temporal relationships among audio , video, and other data ; and the delivery of the data in accordance with these relationships . It i s important to distinguish between liveand synthetic-synchronization . Live-synch, deals with th e presentation of information in the same order as it was originally collected . For synthetic-synch . various pieces of information must be properly ordered and synchronized in time . Based on th e experiences in DiME and HeiTS, we document the implications of using a hybrid or a unified syste m structures on the required techniques for implementing synchronization .

Multimedia synchronization techniques: experiences based on different system structures

Distributed multimedia applications: A review

Multimedia applications comprise several media streams, which are semantically synchronized at different time instants. The application behavior is stored along with the multimedia database using representation mechanisms such as OCPN (object composition Petri nets) or dynamic timed Petri nets (DTPN). It is imperative that one translates the application behavior to the corresponding schedulable entities, such as packets, so that the performance engineering of any system can be done, using the traffic model arising out of the (media related) application behavior as opposed to individual media level behavior. This requires that a function be defined, which takes the stored temporal representation as input and produces packets as output, preserving the semantic relationships among the streams. The authors propose a methodology based on probabilistic, attributed context free grammar (PACFG) to address this issue. They demonstrate the appropriateness of this methodology by applying it to the OCPN/DTPN representation of a typical multimedia application vis-a-vis orchestrated presentation.

Synchronization representation and traffic source modeling in orchestrated presentation

In the last decades, much effort has been spent on the design and provision of sophisticated communication infrastructures. The development of end-user oriented distributed system applications, leaning on top of these communication infrastructures, so far has attracted little attention. This is regrettable, since communication infrastructures can only become useful and profitable if they can be deployed in the context of a sufficient number of distributed applications.
Two important factors determine the success of distributed applications: (1) the provision of high quality application services and protocols at short time scales; and (2) the availability of standards for these services and protocols that can be used for the construction of �open� distributed systems. The achievement of both (1) and (2) can be supported by a suitable design methodology.
A design methodology entails a systematic approach to carry out complex designs, and therefore should incorporate proper concepts that enable the effective structuring of such designs. Concepts currently used for the design and structuring of application protocols appear to be inadequate for this purpose. Also a step-wise design approach that would help to master complexity and shorten development times is currently lacking.
Standards are necessary since individual users of distributed system applications prefer to be independent on any particular manufacturer or vendor when procuring products, while manufacturers prefer to have maximum implementation freedom when developing such products. An �open� protocol standard defines necessary and sufficient conditions for system parts to interact, such that the system parts can be implemented independently of each other.
ISO and ITU-TSS base the development and definition of protocol standards on a �reference model�, called the Reference Model for Open Systems Interconnection (OSIRM). This model comprises a rudimentary form of a design approach and a reference architecture that can be derived with this approach. According to the OSI-RM, the overall application protocol functionality is distributed over three hierarchical protocol layers. Each layer has been assigned a specific functionality, except the highest layer, the Application
Layer, which is made responsible for all remaining protocol functions. Because the functionality of the Application Layer is not delimited it cannot, as opposed to the other layers, be covered by a single protocol standard or a fixed set of protocol standards. Several identified sets of Application Layer protocol functions are defined by separate Application Service Elements (ASEs).
The appropriateness of the OSI-RM for the development and definition of application protocol standards can be criticized on a number of points:
- the reference architecture defined by the OSI-RM is not flexible enough to adequately cope with the diversity of interaction requirements of distributed applications.
- some design concepts are not clearly defined, thus prohibiting their effective application to structuring problems;
- the relationship between high level application requirements and proposed application protocol solutions is unclear;
- the development of application protocol standards generally takes a long time.
This thesis aims at the development of a methodology for the design of application protocols, including application protocol standards, and so addresses the problems mentioned above. The following contributions are made to achieve this aim:
- design quality criteria are proposed that can be used to guide design decisions and to evaluate designs;
- OSI design decisions and design concepts with respect to application protocols are evaluated;
- general-purpose, elementary design concepts are proposed;
- milestones in the application protocol design process are presented;
- behaviour composition and structuring techniques are developed that can be used to represent design results corresponding to the identified milestones;
- design methods are proposed to support the correct performance of design steps between milestones;
- a flexible reference architecture is proposed.
A (potential) result of the design methodology is that layered application protocol hierarchies can be avoided if they are not required by the class of distributed applications that must be supported.

On the design of application protocols

Protocols for high-speed multimedia communications networks

It is often believed that with the mere knowledge of the kind of media, e.g. "speech, the system and specifically the communication requirements of a distributed multimedia application are automatically fixed. In general this assumption is wrong. As a matter of fact, certain technology aspects including the kind of media, coding, and compression techniques are of p k a r y relevance. However the specific application's functionality imposes fixther requirements. In this paper we present a new model for distributed multimedia applications which is used for the description of such applications. It allows for the retrieval of some communication requirements, and it can be used to structure the described application. This model was created within and used for the development of an audio/video distribution service as application of the HeiTS (IBM ENC Heidelberg High-speed Transport System) project in Heidelberg.

R. Steinmetz

Papers

Modelling Distributed Multimedia Applications