No abstract

En este trabajo se propone y evalúa una nueva política de planificación en sistemas multinúcleo de tiempo real. La planificación combinada trata de establecer una nueva estrategia de planificación que reduzca los cambios de contexto y las interferencias generadas con el uso de recursos compartidos entre los núcleos. Esta estrategia se basa en la idea de elegir la mejor política para un determinado intervalo de ejecución durante todo el plan. El resultado es una planificación de tareas implementada con diferentes políticas como EDF (Earliest Deadline First) o DM (Deadline Monotonic) y variantes de las mismas. Se ha evaluado la propuesta con diferente número de núcleos y carga obteniendo mejoras con respecto al uso de un solo planificador en todo el horizonte de ejecución.

In our everyday lives, more and more computers are controlling our environment: mobile phones, industrial processes, driving assistance, etc. All these systems present strict requirements to ensure proper behaviour. In many of these systems, the time at which the action is delivered is as important as the logical result of the computation. About 40 years ago, real-time systems began to attract attention in computing field and nowadays are applied in wide ranging areas as industrial applications, aerospace, telecommunication applications, consumer electronics, etc. Some real-time challenges that must be addressed are determinism and predictability of the temporal behaviour of the system. In this sense, to guarantee program execution and system response times are essential requirements that must be strictly met through appropriate task scheduling strategies.Furthermore, multiprocessor architectures are becoming more popular due to the fact that processing capabilities and computational resources are increasing. A recent study estimates that there is an increasing tendency among multiprocessor architectures to combine different levels of criticality in the same system. In this sense, to provide isolation between applications is extremely required. Partitioned technology is able to deal with this purpose.In addition, energy management is a relevant problem in real-time systems. Many real-time embedded systems, as wearable devices or mobile robots that require batteries, seek to find techniques that reduce the energy consumption and, as a consequence, increase the lifetime of their batteries. Also clear operational, financial, monetary and environmental gains are reached when minimizing energy consumption.Faced with all this, this work addresses the problem of schedulability and contributes to the study of new scheduling techniques in partitioned real-time systems. These techniques provide the minimum time to feasible schedule tasks sets. Moreover, allocation techniques for multicore systems whose main objective is to reduce the energy consumption of the overall system are also proposed.Finally, some of the obtained results are discussed as conclusions and future works are introduced. ResumenEn nuestra vida cotidiana, cada vez más ordenadores controlan nuestro entorno: teléfonos móviles, procesos industriales, asistencia a la conducción, etc. Todos estos sistemas presentan requisitos estrictos para garantizar un comportamiento adecuado. En muchos de estos sistemas, cumplir con las restricciones de tiempo es un factor tan importante como el resultado lógico de los cálculos. Desde hace aproximadamente 40 años, los sistemas en tiempo real son muy atractivos en el campo de la computación y hoy en día se aplican en áreas de gran alcance como aplicaciones industriales, aplicaciones aeroespaciales, telecomunicaciones, electrónica de consumo, etc.Algunos retos a abordar en el campo del tiempo real son el determinismo y la predecibilidad del comportamiento temporal del sistema.En este sentido, garantizar la ejecución del progr...