Real Time Systems: A Super Scheduler Model for Real-Time Systems with capability of Urgent Tasks Scheduling By ANANTHA KRISHNA PILLUTLA K00377332 (11-12.15 BATCH) 1 Abstract- Real time systems mainly work on the principal that how the system normally works in the critical situations. The design of the real time systems is done in this way for the safety critical applications. When all the tasks are already scheduled in the safety critical application, and a critical task arrives in the system, it should be given priority soon after it arrives, otherwise it leads to disaster or a system failure. The main problem in real time system is that we cannot predict the arrival of critical tasks. To address this kind of problems, a scheduler called “super scheduler” is employed. In a hierarchical real time system there are several real time sub systems called “components”. So, in these kind of hierarchical real time system the problem can be more complex. Therefore the super scheduler should work different in different situations accordingly. When the critical task arrives the schedulers preempts all the existing tasks, and re schedules the tasks giving the critical task more priority and the remaining tasks are again prioritized. When the critical task completes its execution, the remaining tasks are executed in the order of their new priorities. The stability of the component is guaranteed by completion of the critical task first and all the noncritical tasks next without missing any of their deadlines. The stability of the hierarchical real time system is guaranteed by the stability of all its individual components. This paper clearly explains the model where all the components including the critical tasks are stabilized and thus the stability of the hierarchical real time system is guaranteed. Moreover, a fault tolerance method has been applied for all components. The evaluation results show that the proposed system improves the stability of the hierarchical real time system by decreasing the list of jobs which miss their deadlines including the critical tasks. Keywords: Super scheduler, Critical task, Real-time Scheduling, Hierarchical real-time systems, Safety-critical applications. I. INTRODUCTION Today, real time systems play a crucial role that affects and controls our lives. A special consideration should be given to the safety critical applications where there systems having critical and vital requirements. A safety critical application is like if the failure of the application occurs it results in disasters like substantial economic loss, extensive environmental damage, or might endanger human life. The main areas which fall under the safety critical applications are automotive, medical care, and industrial automation.In every real time system there is important aspect which has to be kept in mind that is deadline. Deadline is defined a time limit where all the scheduled tasks have to complete their execution, otherwise a disaster might occur. Among all the tasks critical tasks must be given priority and it should be completed within the deadline, if not some serious effect will be there in the safety critical applications. So, the critical tasks must start its execution soon after it enters the system. The main problem in the safety critical applications is that the arrival of the critical tasks in to the system, cannot be predicted. To address this problem we use a scheduler called “super scheduler”. Super Scheduler works in such a way that the critical tasks are prioritized first by changing the priorities of all the tasks in the system. Now, all the tasks are scheduled according to their new priorities. Accordingly the responsibility of the super scheduler is that control and alter the scenario of the existing tasks when the critical task enters. As we cannot predict the critical task arrival, the super scheduler we are using should act according the situation, adaptable and tolerant. These days, real time applications such as electronic equipment used in airplane are more complicated that a simple real time system could not satisfy all the requirements of the application. To survive with the complexity of the present-day real time system, the systems are to be developed individually and then interconnected to form system of systems (SOS). The most popular model in the design of SOS is the hierarchical model. Hierarchical real time system which is the model of system of systems comprises of many real-time sub systems called components. Now, every component in the system will act like simple real time system in which there are many tasks and scheduling algorithms which work individually. Hierarchical real time system will have all its components in different levels. The “root component” is the primary level. The root component in turn will have many components in the next levels. Therefore the components which are present in the middle level of hierarchy will be having a parent component on its top level and many child components in the next level. Components in the same level of the hierarchical model will share the resources which they get from the parent components. In this way the communication between the child and parent components is established.In this paper , for the hierarchical system of systems, a model has been proposed to handle the critical tasks using the super scheduler. When the critical task arrives in o the system, all the components use the super scheduling proposal to schedule the their tasks individually. By using the super scheduler the stability of the system is guaranteed when the critical task suddenly enters the hierarchical real time system. The statbility of the hierarchial real time system depends on the stability of the components when the tasks in the individual components do not miss their deadlines. 2 The remainder of the paper is organized as follows: Section ӀӀ presents an overview of previous studies for scheduling issues in safety-critical real-time systems. In, Section ӀӀӀ to cope with the scheduling problem of critical tasks in hierarchical real-time systems, the proposed super scheduler is described. Section ӀV includes the experimental results and finally conclusions are given in Section V. II. PREVIOUS STUDIES A. Related Work Among of the task scheduling algorithms, the hybrid scheduling model, commonly have proposed for realtime and non-real-time tasks [6], [7]. Hence, a hybrid scheduling model for hard, soft and non-real-time tasks based on two-level hierarchical scheduling architecture is proposed in [8]. In the top-level of this hybrid scheduling model, a constant utilization server is created for each real- time scheduling policy. Moreover, the model has a total bandwidth server for all the non-realtime tasks which schedule by time sharing scheduling policy. The bottom-level scheduler is the operating system scheduler which responsible for scheduling all the servers. This hybrid scheduling is appropriate for the applications that consist of real-time and non-real-time parts. An evaluation of dynamic tasks execution, under scheduling models like hybrid, static and dynamic scheduling is presented in [9]. The evaluation results show that the hybrid scheduling model can simplify selecting the most appropriate design model of the system tasks scheduling. The hybrid scheduling model guarantees that real-time tasks are completed in their deadlines. Hence, this model is appropriate for the realtime systems consist of hard real-time tasks [10], [11]. Moreover, this model can save the real-time system resources. A scheduling method, called fast response time scheduling 1 algorithm is proposed in [12]. This hybrid algorithm is designed to fully take the advantages of Event- based scheduling (EBS) and Run-based scheduling (RBS) algorithms [13], [14]. In comparison to the static scheduling algorithms, using FRTS algorithm can improve the average response time of real-time tasks of the system, significantly. A hybrid model containing static and dynamic methods for real-time tasks scheduling, is proposed in [2]. This hybrid scheduler comprises two phase architectural model. The first phase included an offline scheduler for real-time tasks of the system, called “fixed scheduler”. Moreover, to achieve fault-tolerance for the system, a redundancy method is employed [15]. Finally, in the second phase, the hybrid scheduler is embedded in a scheduling model, called “super scheduler”. To handle critical tasks, the super scheduler alters the priorities of tasks such that the highest priority is given to the critical tasks. Using this scheduling model guarantees the stability of the system, in case of safety-critical applications. In addition to typical scheduling algorithms, the super scheduler model is employed to schedule realtime tasks in arrival of critical tasks. A suitable technique for design of the real-time tasks scheduling is proposed in [3]. Moreover, the method can evaluate the system behavior in case of safety- critical applications. The method is also appropriate for large real-time systems which consist of soft and hard real-time tasks. Thus, using this method can guarantee the stability of a large real-time system, on the arrival of critical tasks. A hierarchical framework contained real- time components are proposed in [5]. Each component of this hierarchical framework is a real-time system which has a set of real-time tasks. These set of tasks can be scheduled with RM or EDF algorithms. Moreover, each component has a fault model for its set of tasks, individually. Under fault conditions of the components tasks, the faults can be detected and recovered [16], [17]. Thus, the task recovery methods, such as task re-execution, and forward/backward recovery can be employed in the components tasks. According to above, the presence of critical tasks is not considered in the hierarchical real-time systems. Further, to use of task recovery methods, any of faults should be detected at the end of a real-time task execution. Hence, to handle critical tasks a faulttolerance hierarchical real-time system that contains the super scheduler can be defined. B. Problem Definition Assume that each component of a hierarchical realtime system has a set of real-time tasks. The problem is studied under overloads conditions of real-time tasks that scheduled by a typical scheduler. The arrival of unpredictable critical tasks is the most common cause of these overloads conditions. These critical tasks will lead to the scheduling scenario changes. Hence, most of the real-time tasks should be complete in their deadlines. Moreover, the stability of the hierarchical real-time system should be guaranteed. III. PROPOSED TECHNIQUE A. System Model The Real time system we are considering is a Hierarchial real time system which is a real time SoS(System of Systems).This contains many different levels of sub systems. Each component of the Real-time system consists of many simple real-time systems, individually each component has several different tasks and a scheduling algorithm. Moreover , each real-time system does concurrent processing with processors to execute the tasks. All the processors have same levels of priority tasks which share the computational resources, which they get from its Parent real-time system. The basic work unit of Hard RealTime Sub System is a component which can be defined by the following : ▪ W : a set of tasks or Workload ▪ R : resource model of the parent processor ▪ A : Scheduling algorithm of workload for the resource model. 3 A workload W set comprises of periodic tasks of real time systems which are executed on the parallel or concurrent processing components. A resource model R tells us the amount of resources allocated from the parent component. In order a periodic resource model is a compositional real time scheduling framework as its global timing properties can be established independently and analysing local timing properties. The two vital issues in growing such a framework are dynamic of the collective real time requirements of a segment undertaking as a solitary constant prerequisite and to make the interest reflection results to the framework level continuous prerequisite. The HRTS has an interface between the parent and the component real-time system. Each parent component take the real-time resources requirements of its child components. The resources required by the child processor is supplied by the parent processor. The interface model is a scheduling framework so that global timing properties can be established by composing independent timing properties. The feasibility condition is guaranteed if the amount of resources is supported by the real time system within the period. The feasibility condition means all the tasks and jobs will be completed within the deadline. IV CONCLUSIONS : Fault tolerance schedulers can be used in critical task for hierarchical real-time systems. Using a super scheduler among schedulers will make sure the stability of a hierarchical real-time system, even when there are critical tasks. For hierarchical real-time system to be stable, all its components should be stable as well. Consequently, using a super scheduler here for each component of a hierarchical real time system guarantees stability of the system. Results on evaluating show that using a super scheduler in a hierarchical real time system, improves the deadline miss ration of tasks to a great percentage. Therefore, employing the proposed super scheduler will improve the stability of hierarchical real-time systems more than typical and hybrid schedulers.. B. Fault Handling Model In the model we are discussing let us assume that each component of the system has a critical manager section. On the arrival of a high priority task the scheduler should handle all the conditions. On the arrival of high priority tasks, the Super scheduler is employed with three phases, the scheduling algorithm schedules the task to each component it schedules the component tasks with a fixed scheduler in which all the tasks are scheduled according to their priorities and they are B. In the second phase, to achieve fault tolerance for the system components, a hardware redundancy method [20] is employed. A primary- backup redundancy is used for the concurrent processors of the system components. The backup copies can be overlapped to reduce the number of backup processors. Moreover, these backup copies of concurrent processors are preserved in the critical manager section of each component. This phase is, called “dynamic scheduler”. In the third phase, the super scheduler is used to schedule the real-time tasks of the components such that, on the arrival of a critical task, the component automatically performs the context switching of fixed scheduler to the super scheduler. This phase is called “critical scheduler”. 4 5