論文

Scheduler 1-2-3 is an interactive analysis tool that can perform schedulability analysis for the development of real-time computing systems. The schedulability analysis can verify whether all hard real-time tasks in a target system will be completed by their deadlines at the system design phase. Scheduler 1-2-3 is a window-based stand-alone tool that can also be used as a synthetic workload generator in an integrated tool set that consists of a timing tool, a schedulability analyzer, and a real-time monitor/debugger.

In this paper, we first address the issues in real-time monitoring and debugging, such as capturing the timing error, the monitor's invasive nature and visualization of system behavior. We then describe the architecture of the ART Real-Time　Monitor which is being built for a distributed real-time OS, called ARTS. The built-in monitoring/debugging kemel primitives are also mentioned.

In the time-driven scheduling (TDS) model, both the importance value and timing criticality of each task are given together as a function of time. The TDS scheduler attempts to maximize the total value of completed tasks. A practical TDS model that has been implemented in the Mach operating system kernel is described. Some of the important design considerations for a flexible TDS scheduler are discussed, with emphasis on policy/mechanism separation as a system structuring principle. The structure of a TDS scheduler that was actually implemented is explained. The practicality of the TDS scheduler is demonstrated by performance results.

In this paper, we describe the critical issues in providing a large grained parallel processing capability based on a network of powerful personal workstations. In particular, we address a policy problem where how an owner of the workstation can express 'willingness' for sharing. We also discuss a 'software' tastbed approach by which a program designer can improve construction of distributed software.

Process scheduling in real-time systems has almost invariably used one or more of three algorithms, namely, fixed priority, FIFO or round robin. The reasons for these choices are simplicity and speed in the operating system, but the cost to the system in terms of reliability and maintainability have not generally been assessed. The notion that the primary distinguishing characteristic of a real-time system is the concept that completion of a process or a set of processes has a value to the system which can be expressed as a function of time is discussed. This notion is described in terms of a time-driven scheduling model for real-time operating systems and provides a tool for measuring the effectiveness of most of the currently used process schedulers in real-time systems. Applying this model, a multiprocessor real-time system simulator was constructed with which a numberf well-known scheduling algorithms such as shortest process time, deadline, shortest slack time, FIFO and a fixed priority scheduler are measured with respect to the resulting total system values. This approach to measuring the process scheduling effectiveness is a first step in the longer term effort to produce a scheduler which will explicitly schedule real-time processes in such a way that their execution times maximize their collective value to the system, either in a shared-memory multiprocessing environment or in multiple nodes of a distributed processing environment.

In an interactive multimedia environment that handles multiple sessions dynamically, a mechanism to control QOS among sessions is very important. As such a mechanism, the QOS-Control Server has been developed on RT-Mach. This paper gives the results of several experiments with this server, and describes the extensions of RT-Mach.

This paper describes the design and implementation of a comprehensive operating system (COS) for highly parallel machine. This is one of the research themes in the Japanese University Highly Parallel Research Project supported by the Special Grant of the Research Funds of Ministry of Education. The target machine is called JUMP-1 originally developed in this project. Several design policies are discussed, and the user service classes are defined. COS architecture is based on the micro kernel, and it includes both Partition OS and Service OS. The development environment is also discussed briefly

Continuous-media applications require more efficient and flexible support from real-time threads than traditional real-time systems. It includes functionalities such as the dynamic management of thread attributes and the support of multiple thread models. In this paper, we will describe the design and implementation of user-level real-time threads on the RT-Mach micro kernel. Since they are implemented at user-level, both of the fast management of thread attributes and the support of multiple thread models are possible

Multimedia applications have timing requirements that cannot generally be satisfied using the time-sharing scheduling algorithms of general purpose operating systems. Our approach is to provide the predictability of real-time systems while retaining the flexibility of a timesharing system. We designed a processor capacity reservation mechanism that isolates programs from the timing and execution characteristics of other programs in the same way that a memory protection system isolates them from outside memory accesses. In this paper, we describe a scheduling framework that supports reservation and admission control, and we introduce a novel reserve abstraction, specifically designed for the microkernel architecture, for measuring and controlling processor usage. We have implemented processor capacity reserves in Real-Time Mach, and we describe the performance of our system on several types of applications.

Synchronization and communication are two common sources of priority inversion which may make the behavior of systems unpredictable and unanalyzable. In microkernel-based systems, they are heavily used for building operating system servers and decomposing applications into several tasks. The management of priorities in IPC and synchronization should be integrated using a uniform mechanism since priority inversion occurs if such integration is not supported. Also, a highly preemptable server structure should be provided because the execution in a server may take up too much time. In this paper, we propose an integrated real-time resource management model, and a real-time server model which solve the above problems. We implemented and evaluated the models in Real-Time Mach using a uniform mechanism. Our approach enables us to build operating system servers and to decompose applications into several tasks without incurring priority inversion.

This paper presents dynamic QOS control schemes based on real-time threads. A real-time thread is based on an explicit declaration of timing attributes, such as start time, deadline, period, and its deadline handler. Two dynamic QOS control schemes: a self-stabilization scheme and a QOS manager-based scheme are evaluated using a QuickTime movie player running on Real-Time Mach

This paper briefly motivates and describes processor capacity reserves and a prototype implementation of reserves in Real Time Mach. The reservation mechanism provides a way for application programs to specify their reservation requests and incorporates a scheduling framework which supports an admission control policy. The reservation system accurately measures processor usage of individual threads, even when these threads invoke user-level servers to do work on their behalf, and the scheduler enforces reservations by relegating reserves to a time-sharing pool after their reservations have been consumed.

Interprocess communication (IPC) provides the fundamental mechanism for the Mach microkernel to be extensible and flexible. Mach IPC provides efficient communication mechanisms for many applications. However, it does not provide sufficient functionality for real-time applications which have rigid timing constraints among threads. In Real-Time Mach (RT-Mach), we have extended Mach IPC to be priority inversion free for real-time applications. This paper describes the Real-Time IPC (RT-IPC) facilities, its implementation, and the evaluation results. We used the Distributed Hartstone (DHS) real-time benchmark for the evaluation and the results show that RT-IPC can reduce priority inversion and improve CPU utilization for real-time applications.

In multimedia systems, session coordination technique is important to accommodate a new session and re-distribute reclaimed resources when tough continuous media session is exiting. Although end-users do not want to specify the quality of service (QOS) in an exact memory size like in KB or KB/s of bandwidth, such rigid usages are necessary for admission control and coordination. Therefore, we introduce the multi-level quality specification and QOS translation mechanism to put it into the quantitative expression which the system can handle. Nonetheless, session coodination have demanded complicated computation or presented an NP-complete problem. Constraining QOS specification, we invent a simplified session coordination method which is feasible enough to actually implement. This paper reports experiments on QOS translation and multiple session coordination using the Conductor/Performer system on top of Real-Time Mach 3.0.

Java has many benefits such as security in distributed environments, reusability of code, and portability because it is architecture neutral. From those characteristics, Java is beginning to be used in many environments where more popular languages such as C and C＋＋ were originally used. Even though Java provides benefits, still it has problems which must be overcomed. There are often real-time constraints that should be met in those applications. Current Java execution environment and language specification cannot satisfy those requirements. Our research focuses on the current limits of the Java language and its execution environement, and to seek the posibility for real-time using Java and the technologies we have available today. To investigate issues in real-time Java, we have implemented a prototype real-time Java environment and Real-Time Java thread.

In an interactive multimedia environment that handles multiple sessions dynamically, a mechanism for controlling the QOS among sessions is very important. It is, however, difficult to achieve such QOS control without any system support. A new resourcemanagement mechanism and QOS-control architecture for multimedia applications are needed. This paper first discusses features required for such architecture, and then proposes a new QOS-control architecture based on ''QOS-Ticket.'' In this QOS-Ticket model, resource management and QOS control are achieved through the cooperation of an operating system, a QOS Manager, and individual continuous-media sessions.'