Research On Some Key Technologies Of The Operating System Supporting Mobile Devices In Ubiquitous Computing Envrionment

A brand new computing model - Ubiquitous Computing, emerges from academic papers to actual applications with the development of computing technologies. "Immanence" is the key character of Ubiquitous Computing which means there are embedded, mobile or fixed computing or sensor devices around every visible or invisible corner. Operating system, which is fundamental system software, plays an important role in managing hardware and software resources along with providing users with convenient Human-Computer Interfaces. Even if the concept of Ubiquitous Computing has been brought out for about a decade there are still many issues to be explored, especially supporting software. Besides many kinds of embedded operating systems, devices in ubiquitous computing environment have other special requirements, such as context-awareness, self-adaptation, reliability, etc, that generates a pressing need for designing specific operating systems to support those applications. Mobile devices are typical instruments in ubiquitous computing environments. Its processing power is between embedded processors and personal computers and can embody the features of Ubiquitous Computing thoroughly. Therefore, it is very meaningful to research key technologies of operating systems which support their work. In this dissertation, we stress our research on some key technologies of the operating system which supports mobile devices in ubiquitous computing environment, from three aspects, i.e. schedulibility, reliability and reconfigurability. The purpose of the dissertation is to bring forward some new models and algorithms and set up the development foundation of ubiquitous operating systems.The main contents and contributions of the work can be summarized as follows:1. The system infrastructure of Ubi-OS is proposed. Functionalities of interrupt scheduling, network scheduling, reconfigurable hardware tasks and on-line software maintenance are added or improved according to computing requirements of mobile devices in ubiquitous computing environments. Furthermore, energy-efficient FLASH file system, real-time memory management and priority based weighted round-robin task scheduling algorithm are designed and implemented as well. The operating system is based on micro-kernel. Other function modules like Inter Process Communication, device management, are gotten from COTS (Commercial Off the Shelf) operating systems directly which ensures the performance and stability.2. The novel interrupt management mechanism which supports system reliability and interaction is designed and implemented. Interrupts is one of the important ways through which operating systems interact with outside environment. The effects of interrupt management seriously affect the reliability, real-time capability and other performance of the system. Interrupt sharing and interrupt scheduling are added to interrupt management. Interrupt sharing help the system make use of more I/O devices under the limitation of hardware resource. A scheduling algorithm called LEDF is proposed to schedule interrupts, which could decrease interrupt overtiming when frequent interrupts happening. The overhead is very low when there is no need for scheduling and can be accepted compared to overtime interrupts.3. Aiming at various network environments which mobile devices will encounter when moving, a network scheduling scheme related to operating system, is brought forward, together with a scheduling algorithm. Necessary data structures are designed. Seamless data transforming is explored to some distance. The scheduling scheme could change the service data flow from one network to another, in terms of the requirement of context.4. TVIA algorithm, which is developed from KVIT algorithm, is presented to fetch up the deficiency of current placement methodology for reconfigurable hardware tasks. The efficiency and success rate of task placement is enhanced by supporting task transforming and adding reasonable data structures. New arrived tasks will be placed adjacency to the vertexes of already running tasks. By this approach, the wasted space is minimized, and the chip utilization is maximized. Five strategies for reusing configuration of hardware tasks, which try to increase hitting rate by various means, are proposed. Times of reconfiguration can be greatly decreased via reusing configuration. The scheduling of hardware tasks is a critical factor to affect the performance of dynamic reconfigurable systems. The proposed reuse based earliest last start time first (R-ELSTF) algorithm adopts TVIA as placement method, adopts Best-Fit as reusing configuration method and ELSTF as scheduling method. According to experiments, the algorithm has better scheduling quality, low overhead and supports real-time and none real-time simultaneously.5. Self-adaptive on-line software maintenance scheme is proposed to implement software reconfiguration. The system framework of the scheme, hiberarchy and concrete methods are designed and implemented. Mobile devices often have limited computing resources such as computing power and memory whereas they are required to provide users with various services in terms of different context. The conflict lies in both hardware and software. Mobile devices can not store as many software as possible. The software running or stored on devices should be changed complying with services dynamically. The proposed SOSM method could solve the problem through three scales that are data and codes, components, applications. Applications software, which are developed conforming to SOSM method, can be easy maintained by Ubi-OS.Based on the above studies in this dissertation, we design and implement interrupt scheduling algorithm, framework and structure of OS network scheduling, placement algorithm of hardware tasks, configuration reuse strategies for hardware tasks, real-time scheduling algorithm for hardware tasks, methods and models of self-adaptation on-line software maintenance. The experimental results show that these key technologies, which are proposed from three aspects, can improve self-adaptation capability, intelligence, efficiency and reliability of the system. They can keep down complexity of application software design as well.