Design and implementation of a single system image operating system for ad hoc networks

Hongzhou Liu, Tom Roeder, Kevin Walsh, Rimon Barr, Emin Gün Sirer
2005 Proceedings of the 3rd international conference on Mobile systems, applications, and services - MobiSys '05  
In this paper, we describe the design and implementation of a distributed operating system for ad hoc networks. Our system simplifies the programming of ad hoc networks and extends total system lifetime by making the entire network appear as a single virtual machine. It automatically and transparently partitions applications into components and dynamically finds them a placement on nodes within the network to reduce energy consumption and to increase system longevity. This paper describes our
more » ... ogramming model, outlines the design and implementation of our system and examines the energy efficiency of our approach through extensive simulations as well as validation of a deployment on a physical testbed. We evaluate practical, power-aware, general-purpose algorithms for component placement and migration, and demonstrate that they can significantly increase system longevity by effectively distributing energy consumption and avoiding hotspots. Efficiency: The system should execute distributed ad hoc network applications in a manner that conserves power and extends system lifetime. Policies and mechanisms used for adaptation in the systems layer should not require excessive communication or power consumption. Adaptation: The system should respond automatically to significant changes in network topology, resource availability, and the communication pattern of the applications. Adaptation should not require a priori involvement from the application programmer. Generality: The system should support a wide range of applications. Developing new applications that execute efficiently on an ad hoc network should require little effort. The system should provide effective default adaptation policies for applications that are not power-aware. Applications should be able to direct and, when executed with sufficient privilege, override the default adaptation using application-specific information. Extensibility: The system should provide facilities for deploying, managing and modifying executing applications whose lifetime may exceed those of the network participants. Compatibility: The system should not require mastering a new paradigm in order to deploy applications. Standard development tools should continue to work in building applications for ad hoc networks. The system should enable applications to execute on ad hoc networks of heterogeneous nodes. MagnetOS meets these goals by making the entire network operate as an extended Java virtual machine. Mag-netOS applications are structured as a set of interconnected, mobile event handlers, specified statically by the programmer as objects in an object-oriented system. The MagnetOS runtime uses application partitioning and dynamic migration to distribute the event handlers to nodes in the ad hoc network and finds an energy-efficient placement of handlers. MagnetOS applications are comprised of event handlers that communicate with each other by raising well-typed events. Event signatures specify the types of the arguments passed with the event, as well as the return type of the handler. By default, all externally visible entry points, such as methods in a Java object specification, serve as event declarations, and method bodies constitute the default handler for that event in the absence of overriding runtime event bindings. Consequently, the MagnetOS programming model closely parallels the Java virtual machine, providing access to standard Java libraries and enabling familiar development tools to be used to construct distributed applications. Our MagnetOS implementation consists of a static application partitioning service that resides on border hosts capable of injecting new code into the network, a runtime on each node that performs dynamic monitoring and component migration and a set of policies to guide object Figure 1 : A static partitioning service converts monolithic Java applications into distributed applications that can run on an ad hoc network and transparently communicate by raising events.
doi:10.1145/1067170.1067187 dblp:conf/mobisys/LiuRWBS05 fatcat:louevdu2abfynipeepflrjt63i