No abstract

No abstract

PERTS is a prototyping environment for real-time systems. It is being built incrementally and will contain basic building blocks of operating systems for time-critical applications, tools and performance models for the analysis, evaluation and measurement of real-time systems, and a simulation/emulation environment.It is designed to support the use and evaluation of new design approaches, experimentations with alternative system building blocks, and the analysis and performance profiling of prototype real-time systems.

PERTS is a prototyping environment for real-time systems. It is being built incrementally and will contain basic building blocks of operating systems for time-critical applications, tools and performance models for the analysis, evaluation and measurement of real-time systems, and a simulation/emulation environment.It is designed to support the use and evaluation of new design approaches, experimentations with alternative system building blocks, and the analysis and performance profiling of prototype real-time systems. I. INTRODUCTION While existing approaches, techniques and tools for the design, prototyping and development of software systems are effective for many application domains, they often do not address the difficulties in building hard real-time computing systems. A hard real-time computing system, hereafter simply called a real-time system, is one in which most tasks have hard timing constraints. Here, the term task refers to a basic unit of work. A task may be a granule of computation, a unit of data transmission, a file access, or an I/O operation, etc. The simplest timing constraint imposed on a task is its deadline, the point in time by which the task is required to complete. The result produced by a task with a deadline is correct only if it is available by the deadline, in addition to being functionally correct. A late result is of little or no use. Applications supported by real-time systems include command and control, guidance and navigation, flight control, object identification, autonomous vehicle control, and intelligent manufacturing.approach would begin with models and optimality criteria that explicitly account for the constraints and possibilities of trading off between various figures of merit, and then design the application and the underlying system to achieve the desired tradeoffs. Such an approach would lead to more flexible, easy-)_ This work has been partially supported by ONR Contract Nos. NOOO14-89-J-1181, and N000-92-J-1815. 372L https://ntrs.nasa.gov/search.

In this paper, we propose an eflcient algorithm t o statically assign tasks and fogical resources t o the processors, taking into account the blocking times due to resource contention. W e formulate the assignment problem using an assignment graph where the weights on the edges of the graph represent the desirability of assigning the tasks (connected b y the edges) to the processors. These weights are computed based on the resource usages, communication costs, and the priorities of the tasks. A n eficient clustering and graphbisection algorithm is used to find an assignment of the tasks and resources such that the total communication costs and total blocking time are minimized, subject to the constraint that the tasks are schedulable after the assignment. Ezperimental results indicate the effectiveness of our algorithm.43 0-8186-6420-7194 $03.00 0 1994 IEEE