Description

Multimedia metacomputing is a new approach to the management and processing of multimedia data in web-based information systems. It offers high flexibility and openness while shielding the applications from any system internals. This is achieved by realizing several abstraction concepts like device independence, location transparency, execution transparency, and data independence. However, format independence as an important aspect of data independence is well known to be fairly hard to realize, because it tends to reduce performance massively and, additionally, may lead to unexpected data loss.

To solve these problems, we introduce an advanced abstraction concept called transformation independence, which includes all concepts mentioned above, but adds some more concepts suitable for the metacomputing approach:

• a flexible materialization management,
• abstract, freely combinable media filters as the basic operators for metacomputing, and
• the concept of transformation requests, which express specific application semantics.

In the VirtualMedia project, a formal model for specifying complex media operations as transformation requests and for processing and optimizing such requests is being developed. The model is based on the well-known filter graph approach which is extended by so-called magic channels. A magic channel implicitly describes a set of media operations on the media data that is sent through the channel. Hence, magic channels are appropriate to create and represent transformation requests. The model also provides a request processing mechanism, which allows to eliminate all magic channels from a given request graph, thus finding an optimal plan for executing the transformation request.

We define a generic VirtualMedia architecture consisting of two major components:

• request processing, which is built by sub-components for request translation, graph transformation, and materialization management, and
• instantiation service, which is built by sub-components for cost-based optimization and resource management.

A concrete VirtualMedia system can be built upon component-based middleware, which principally enables an Internet-spanning collaboration of media servers, media clients, and media processors. However, this requires the definition of open standard interfaces and protocols allowing autonomous service providers to implement and integrate media management and processing services independently. Another architecture approach is based on the extensibility of modern object-relational DBMS. This feature allows to integrate multimedia metacomputing facilities directly into ORDBS, which is, e. g., a promising perspective for the improvement of digital libraries.

Compared to traditional ADT-based approaches the VirtualMedia model offers considerably more optimization potential, extensibility, abstraction etc.. Although there is no implementation, yet, first validation indicates the effectiveness of the proposed heuristics for request processing, which already yield good results by using a simple greedy algorithm. More sophisticated algorithms like simulated annealing or evolutionary algorithms are also applicable, which may result in even better execution plans.