Study of time characteristics of subtasks in real-time multiprocessor systems

Abstract

The article examines the time characteristics of subtasks in real-time multiprocessor systems. It is noted that real-time systems are systems in which, along with the logical correctness of the computing process, it is also necessary to ensure the timeliness of its flow. It is emphasized that the entire range of functionality is presented as a set of simpler tasks, for the execution of which processor elements are selected. When forming each processor element, planning discipline is mandatory, according to which the time requirements of the tasks are analyzed. It is emphasized that the organization of calculations takes into account the time spread of the execution time, in which tasks with a large spread are divided into two subtasks, the first of which is completed in a time that does not exceed the average time of the original task, and the second represents the remainder of its execution. The main task directed to execution is mathematically formed and the algorithm for dividing it into subtasks is given. The first subtask is scheduled similarly to other tasks with deterministic characteristics, and the second is queued to the sporadic server. It is shown that with the help of such transformations it is possible to achieve a higher efficiency of the use of processors. It is emphasized that each task can be represented by a set of some elementary blocks, each of which is described by its execution time profile. A lemma is proved for a set of tasks that consists of sporadic tasks with a limited execution time and can be scheduled on several identical processors for any job-saving algorithm. It is noted that there are two main sources of obstacles in the performance of the subtask. In order for any subtask to be scheduled, its execution time must be sufficient to complete it plus the workload of obstacles. It is emphasized that the main condition for successful planning of parallel execution of subtasks is the formation of a number of additional local parameters