Study On Multi-Kernel Component-Based Embedded Operating System

Along with the progress of electronic technology, the software and hardware resources are increasingly integrated into embedded systems. The field of embedded systems is itself undergoing dramatic change, from a field dominated by electrical and mechanical considerations to one that far more closely resembles traditional computing. Embedded operating systems need to provide much more than traditional RTOS functionality, e.g. file system, network protocol and graphical user interface for customers' needs. It's most desirable that popular applications such as high performance multimedia and signal processing are optimally supported. Furthermore, security problems become much important for networked devices.These requirements pose very significant challenges for development of embedded operating system, including extensibility, real-time-ness, general-ness, effectivity and security. Some solutions have been presented for these diverse requreiments. e.g. component technology addressing extensibility problem and dual-kernel technology addressing coexistence problem of real-time-ness and general-ness.Component technology is a popular means for creating complex software system by assembling off-the-shell building blocks. Using component-based development as a design methodology for providing configurable operating systems is a promising direction. However, general component technologies have issues of real-time performance, size and cost which are critical for embedded operating systems. The following key issues need to be addressed by component-based embedded operating system research: guaranteeing effectivity of component-based runtime environment;guaranteeing optimal support for important application fields including multimedia and signal processing;guaranteeing security, real-time-ness and general-ness.This thesis focuses on the study on component-based embedded operating system. It presented a multi-kernel operating system technology. And it designed and implemented a component-based embedded operating system, called Pcanel, along with a component model and a dataflow framework. Its contributions and novelties are mainly in the following aspects:1) We presented a multi-kernel operating system technology (for short, multi-kernel technology): an operating system consists of multiple kernels and a set of components implementing specific functionalities and suppors diverse applications. There are two classes of components according to their structures. The first one is called function component whose executable code implementing all functionalities is self-contained at runtime. The other is called tower component which dynamically loads executable code at runtime. And there are four classes of kernels according to their roles. The first one supports general communication and cooperation of components. The second one supports high performance data transfer and scheduling of components. The third one supports computational resource partitioning of tower components. The last one is standardized operating system kernel in the form of tower component. We designed a component-based embedded operating system using multi-kernel technology, called Pcanel. It consists of five kernels that are C-Kernel supporting component-based runtime environment, D-Kernel supporting high performance data-flow processing, V-Kernel supporting virtualization platform SmartVP, as well as standard real-time operating system T-Kernel and general-purpose operating system Linux running on SmartVP.2) We implemented a component model SmartCM achieving both effectivity of component-based runtime environment and extensibility of advanced features. SmartCM is supported by C-Kernel which has a micro-kernel structure designed for real-time responsibility and compact functionalities including thread scheduling, address space management and synchronous message passing. Based on standard interface definition language, fast remote method invocation is implemented by optimizing the interaction mechanisms of protected components. User applications can extend and customize the component model by exploiting reflective mechanisms. Compared to general component models. SmartCM significantly improves the performance of invoking components and is adequate for embedded systems.3) We implemented a dataflow framework SmartDF supporting application fields of high performance multimedia and signal processing based on D-Kernel. The dataflow framework implements a concurrent model of computation, and provides a dataflow scheduling based runtime environment. A component offers a set of elements to SmartDF. Each element contains a set of computing actions and interacts with runtime environment through a set of ports. Thus elements are loosely coupled and easily reusable. The composition of elements is under canonical models of computation such as state-chart, discrete event and synchronous dataflow. By managing the communication and actions of elements, SmartDF supports dataflow processing applications effectively.4) We extended dataflow framework SmartDF to support multiprocessor System-on-Chip (MPSoC) architectures. There is a trend towards MPSoC for embedded hardware platform. SmartDF provides a distributed runtime environment which suits well such architectures. To support concurrent periodic real-time dataflow procesing on MPSoC, SmartDF proposes a real-time dataflow scheduling model to turn preemptive scheduling of real-time tasks into scheduling of data flows, which is of tremendous practical value by avoiding heavy overhead of context switch.