Forschung : Projekte : NEST

Time-Based Coordination of Networked Embedded Nodes

DARPA Networked Embedded Software Technology (NEST)
In Cooperation with the DREAM Laboratory, University of California, Irvine


Large-scale networks of computer-embedded communication-capable devices are expected to emerge as a major type of computer-based application system structures in not too distant future. One of the major technological challenges that engineering of such systems poses is to facilitate such computer-embedded devices to cooperate in expected and dependable manners all the time while the connectivity among them as well as their locations changes dynamically. If such cooperating actions are subject to timing constraints, whether the constraints are on action times or finishing times, the technological challenge becomes a severe one.

Current commercial operating systems such as Windows-based systems enable relatively simple interactions among programs running on multiple computing nodes networked via both cables and/or wireless communication channels. A collection of programs running on multiple computing nodes toward achieving common application objectives is called a distributed computing program. A layer of software placed on top of such operating systems and distributed over multiple computing nodes to enable easy programming of distributed computing application programs involving sophisticated interactions among their distributed program-segments, is called a middleware. The project is aimed to establish a middleware architecture and prototype that enable large-scale networked embedded computing application developers to assure in cost-effective manners that their application products possess sufficient degrees of three desirable properties, timeliness, fault-tolerance, and security, under variable conditions of functional load, node failures, and network topology. One of the new fundamental design paradigms being explored here is the time-based coordination of distributed actions that is provided by the TMO concept of Prof. K. Kim’s research team at UC Irvine. TMOs can potentially enable much easier and more reliable integration and coordination of various middleware components than other existing approaches do.

Objectives
The problem is that any implementation of a middleware-layer relies on the timing behavior of the underlying operating system. As long as this lower software layer does not provide fully predictable resource access, the designer of a real-time application finally relies on an in-depth knowledge about the specifics of the operating system and intensive monitoring and testing in order to implement the objects with the specified timing. And in this case the purely local timing behavior of an object, as expressed by timing annotations, is not sufficient when the overall system behavior caused by interactions between different object has to be considered. Thus, an abstraction for the monitoring of end-to-end behavior has to be introduced. It is the goal of this cooperation to provide a set of tools, that can be easily deployed in an existing large scale distributed object-oriented system and that provides visual and intuitive feedback on the systems timing behavior, with the option to go into fine-grained details of the real-time computation, such as task scheduling and interrupt processing.

Approach
The approach to give a better view onto system’s behavior is to exploit the activity concept for the purpose of monitoring in the context of distributed object-oriented real-time systems. In order achieve rapid success, the existing monitoring tools of the research group on Real-Time Systems and Communication (Echtzeitsysteme und Kommunikation) in Magdeburg will be adapted to the target environment and the optional system-level instrumentation will be applied to the TMO infrastructure.

Work Plan

• Adaptation of the system specific instrumentation to the projects common platform operating system
·
• Identification of the specific requirements of the TMO object-model, that need to be incorporated into the generic measurement and visualization model
·
• Sensor and agent integration
·
• Implementation of the extended system visualization
·
• Definition of a representative large scale scenario that can used to verify the design assumptions
·
• Experiments in the defined scenario

<< zurück