Evaluation of A Robust Middleware for Numerous Distributed Task-Handling Masayuki Iwai 1 and Hideyuki Tokuda 21 1 Graduate School of Media and Governance, Keio University 2 Faculty of Environmental Information, Keio University 5322 Endo, Fujisawa, Kanagawa 252–8520, Japan E-mail: {tailor, hxt}@ht.sfc.keio.ac.jp Abstract A ubiquitous computing environment entails numerous embedded devices with enough computational power. Such distributed computing resources can cooperate to create ap- plications which support users’ everyday lives. Such an environment requires robust middleware that can handle a large quantity of messages, such as event data or sensor data are occurred between devices. However, the lack of middleware technologies with such functionalities has been making it difficult to create ubiquitous computing applica- tions that we can depend on in their everyday life. In this paper, we propose a novel middleware, named RT-Dragon, which can handle numerous messages with robustness. RT-Dragon can work effectively on systems both real- time and non-realtime runtime because of the mechanism of two level priority-based task processing. In this paper additionally, we evaluate the effectiveness of RT-Dragon on non-RT runtime. 1. Introduction In recent years, embedded devices and tiny sensor de- vices with a network interface have been rapidly evolved. Gumstick-size[2] Linux, PDA, and cellular phone are hav- ing enough processing power both to run simple programs and to communicate to other devices via a network inter- face. These equipments around users receive users’ request and generate messages to other devices. This will cause network traffic incensement. Smart sensors, especially, like MICA Mote[1] and SmartIts[4], can be placed in our liv- ing space. These devices can gather large amount of real world sensor data constantly. These sensor technologies ac- celerate to increase quantity of data. It is an emerging is- sue in ubiquitous computing environment to handle numer- ous messages from lots of embedded devices and sensors. For each message, there are numerical values for example a room temperature value. So it is needed to calculate these values consuming CPU power at a moment. As a result, nu- merous incoming data from distributed devices lead to fail to handle all of messages. To protect failure of components, we propose a two-level priority-based task queuing mecha- nism. This mechanism separates task queuing into two pri- ority tasks. One is the high priority level task and the other is the low priority level task. High priority task is used when the task is sure to do. For example, a networked infrared de- tector of thieves must deliver an emergency task to an elec- tric door to lock it immediately. Such systems need to notify user’s coming task as soon as possible, no matter how the system is overloaded. Without a good designing system, a system programmer would have to pay attention to these complicated issues. We implemented two level priority- based task queuing as a middleware, named RT-Dragon. RT-Dragon is one of the facilities of Dragon. Dragon[8, 6] is a system created on top of [9] and Java that can easily, im- provisatory configure relationships among events generated at distributed components. The remainder of this paper is organized as follows. In section 2 we discus about issues on numerous message handling on distributed environment. In section 3, we explain priority-based task handling middle- ware architecture, referred as RT-Dragon. The evaluation of RT-Dragon is mentioned in section 4, related works are addressed in section 5, and summary in section 6. 2. Issues on Message Handling In this section, we discuss the requirements in realizing a middleware technology to handle numerous message tech- nology to handle numerous message in a ubiquitous com- puting environment. To simplify this discussion, we use the term of Component. Representing a software entity that wraps a sensor or device in ubiquitous computing environ- ment. A component sends event messages, including sensor 1 Proceedings of the 11th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA’05) 1533-2306/05 $20.00 © 2005 IEEE