Dynamic resource management is a difficult problem in modern microservice applications. Many proposed methods rely on the availability of an analytical performance model, often based on queueing theory. Such models can always be hand-crafted, but this takes time and requires expert knowledge. Various methods have been proposed that can automatically extract models from logs or tracing data. However, they are often intricate, requiring off-line stages and advanced algorithms for retrieving the service-time distributions. Furthermore, the resulting models can be complex and unsuitable for online evaluation. Aiming for simplicity, we in this paper introduce a general queuing network model for microservice applications that can be (i) quickly and accurately solved using a refined mean-field fluid model and (ii) completely extracted at runtime in a distributed fashion from common local tracing data at each service. The fit of the model and the prediction accuracies under system perturbations are evaluated in a cloud-based microservice application and are found to be accurate.

The policy used to implement a control algorithm in a real-time system can significantly affect the quality of control. In this paper, we present a method to adapt the controller implementation, with the objective to improve the system’s performance under real-time faults. Our method compensates for missing state updates by adapting the controller parameters according to the number of consecutively missed deadlines. It extends the state-of-the-art by considering dynamic controllers, which have had limited coverage in previous literature. The adaptation mechanism can be precomputed offline, solely based on knowledge about the controller and not on the controlled plant. The approach is indifferent to the control design, as well as to the scheduling policy, and can be automatically realised by the operating system, thus improving the robustness of the control system to intermittent and unexpected real-time faults. We develop a stochastic performance analysis method and apply it to both a real plant and numerous simulated plants to evaluate our adaptive controller. Complementary to the stochastic analysis, we also do worst-case stability analysis of the resulting system. The results confirm the conjuncture that the adaptive controller improves both the performance and robustness in the presence of deadline misses.

Control systems are by design robust to various disturbances, ranging from noise to unmodelled dynamics. Recent work on the weakly hard model - applied to controllers - has shown that control tasks can also be inherently robust to deadline misses. However, existing exact analyses are limited to the stability of the closed-loop system. In this paper we show that stability is important but cannot be the only factor to determine whether the behaviour of a system is acceptable also under deadline misses. We focus on systems that experience bursts of deadline misses and on their recovery to normal operation. We apply the resulting comprehensive analysis (that includes both stability and performance) to a Furuta pendulum, comparing simulated data and data obtained with the real plant. We further evaluate our analysis using a benchmark set composed of 133 systems, which is considered representative of industrial control plants. Our results show the handling of the control signal is an extremely important factor in the performance degradation that the controller experiences - a clear indication that only a stability test does not give enough indication about the robustness to deadline misses.

Resource management in cloud computing is a difficult problem, as one is often tasked with balancing between adequate service to clients and cost minimization in dynamic environments of many interconnected components. To make correct decisions in these environments, good performance models are necessary. A common modeling methodology is to use networks of queues, but as these are prohibitively expensive to evaluate for many real-time applications, different approximation methods for important metrics are frequently employed. One such method—that provides both transient solutions and short, scalable computation times—is the fluid model, which approximates the dynamics of the mean queue lengths using a system of ordinary differential equations. However, finding a fluid model that can adequately approximate an arbitrary queueing network is in general difficult. In this paper, we extend the state of the art with the following three contributions. First, we show that for any mixed multiclass queueing network of processor sharing and delay queues with phase-type service time distributions, such a fluid model can be found via the mean-field approximation. Furthermore, we propose an improved model based on smoothing of the processor share function that improves the performance of certain systems. Finally, using the smoothed mean-field model, we introduce an accurate closed-form approximation of the response time CDF over any subset of classes and queues. The contributions are further evaluated in a large simulation experiment, which shows that they can be used to accurately predict performance metrics under some system perturbations common in cloud computing.

The real-time implementation of periodic controllers requires solving a co-design problem, in which the choice of the controller sampling period is a crucial element. Classic design techniques limit the period exploration to safe values, that guarantee the correct execution of the controller alongside the remaining real-time load, i.e., ensuring that the controller worst-case response time does not exceed its deadline. This paper presents DMAC: the first formally-grounded controller design strategy that explores shorter periods, thus explicitly taking into account the possibility of missing deadlines. The design leverages information about the probability that specific sub-sequences of deadline misses are experienced. The result is a fixed controller that on average works as the ideal clairvoyant time-varying controller that knows future deadline hits and misses. We obtain a safe estimate of the hit and miss events using the scenario theory, that allows us to provide probabilistic guarantees. The paper analyzes controllers implemented using the Logical Execution Time paradigm and three different strategies to handle deadline miss events: killing the job, letting the job continue but skipping the next activation, and letting the job continue using a limited queue of jobs. Experimental results show that our design proposal - i.e., exploring the space where deadlines can be missed and handled with different strategies - greatly outperforms classical control design techniques.

This paper presents JitterTime, a small Matlab toolbox for calculating the transient performance of a control system in non-ideal timing scenarios. Such scenarios arise in networked and embedded systems, where several applications share a set of limited and varying resources. Technically, the toolbox evaluates the time-varying state covariance of a mixed continuous/discrete linear system driven by white noise. It also integrates a quadratic cost function for the system. The passing of time and the updating of the discrete-time systems are explicitly managed by the user in a simulation run. Since the timing is completely handled by the user, any complex timing scenario can be analyzed, including adaptive scheduling and reconfiguration between different system modes. Three examples of how the toolbox can be used to evaluate the control performance of such time-varying systems are given.

Harmonic periods have wide applicability in industrial real-time systems. Rate monotonic (RM) is able to schedule task sets with harmonic periods up to 100% utilization. Also, if there is no release jitter and execution time variation, RM and EDF generate the same schedule for each instance of a task. As a result, all instances of a task are interfered by the same amount of workload. This property decreases the jitters that happen during sampling and actuation of the tasks, and hence, it increases the quality of service in control systems. In this paper, we consider the problem of optimal period assignment where the periods are constrained to be harmonic and the task set is required to be feasible. We study two variants of this problem. In the first one, the objective is to maximize the system utilization, while in the second one, the goal is to minimize the total weighted sum of the periods. First, we assume that an interval is determined a priori for each task from which its period can be selected. We show that both variants of the problem are (at least) weakly NP-hard. This is shown by reducing the NP-complete number partitioning problem to the mentioned harmonic period assignment problems. Afterwards, we consider a variant of the second problem in which the periods are not restricted to a special interval. We present two approximation algorithms with polynomial-time complexity for this problem and show that the maximum relative error of these algorithms is bounded by a factor of 1.125. Our evaluations show that, on the average, results of the approximation algorithms are very close to an optimal solution.

Harmonic task scheduling has many attractive properties, including a utilization bound of 100% under rate-monotonic scheduling and reduced jitter. At the same time, it places a severe constraint on the task period assignment for any application. In this paper, we explore the use of harmonic task scheduling for applications with multiple feedback control tasks. We present algorithm for finding harmonic task periods: to minimizes the distance from an initial set of non-harmonic periods. We apply the algorithm in a scheduling and control co-design procedure, where the goal is to optimize the total performance of a number of control tasks that share a common computing platform. The procedure is evaluated in simulated randomized examples, where it is shown that, in general, harmonic scheduling combined with release offsets gives better control performance than standard, non-harmonic scheduling.

We consider a real-time system of multiple tasks, each task having a plant to control. The overall quadratic control cost is to be optimized. We exploit the periodicity of the task response time, which corresponds to a periodic delay pattern in the feedback control loop. Perturbed periods are used as a tool to find a finite hyper period. We present an analytical procedure to design a periodic linear-quadratic-Gaussian (LQG) controller for tasks with fixed execution times as well as a numerical solution to the periodic -- stochastic LQG problem for tasks with variable execution times. The controllers are evaluated using simulations in real-time scheduling and control co-design examples.

This work presents an event-based control structure using the generalized predictive control (GPC) algorithm with actuator deadband. The main objective of this work is to limit the number of controlled system updates. In this approach, the controlled process is sampled with a constant sampling time and is updated in an asynchronous way that depends on the obtained control signal value. To achieve the desired control properties, a hybrid system framework is used to model the virtual actuator deadband. The deadband is modeled as system input constraints in the GPC's optimization procedure, exploring mixed integer quadratic programming (MIQP) techniques. The presented control structure considers adjustable actuator deadband as an additional tuning parameter and allows a tradeoff between control performance and the number of events/updates. This fact is very important for the lifetime of the actuators and minimizes their wear, especially for mechanical and electromechanical systems. The minimization of actuator usage can be also understood as an economical saving. Other benefits can be found in distributed control systems, where various control system agents are connected through computer networks, and where a reduction in communication is always welcome.

Many cyber-physical systems comprise several control applications sharing communication and computation resources. The design of such systems requires special attention due to the complex timing behavior that can lead to poor control quality or even instability. The two main requirements of control applications are: (1) robustness and, in particular, stability and (2) high control quality. Although it is essential to guarantee stability and provide a certain degree of robustness even in the worst-case scenario, a design procedure which merely takes the worst-case scenario into consideration can lead to a poor expected (average-case) control quality, since the design is solely tuned to a scenario that occurs very rarely. On the other hand, considering only the expected quality of control does not necessarily provide robustness and stability in the worst-case. Therefore, both the robustness and the expected control quality should be taken into account in the design process. This paper presents an efficient and integrated approach for designing high-quality cyber-physical systems with robustness guarantees.

Many embedded systems comprise several controllers sharing available resources. It is well known that such resource sharing leads to complex timing behavior that can jeopardize stability of control applications, if it is not properly taken into account in the design process, e.g., mapping and scheduling. As opposed to hard real-time systems where meeting the deadline is a critical requirement, control applications do not enforce hard deadlines. Therefore, the traditional real-time analysis approaches are not readily applicable to control applications. Rather, in the context of control applications, stability is often the main requirement to be guaranteed, and can be expressed as the amount of delay and jitter a control application can tolerate. The nominal delay and response-time jitter can be regarded as the two main factors which relate the real-time aspects of a system to control performance and stability. Therefore, it is important to analyze the impact of variations in scheduling parameters, i.e., period and priority, on the nominal delay and response-time jitter and, ultimately, on stability. Based on such an analysis, we address, in this paper, priority assignment and sensitivity analysis problems for control applications considering stability as the main requirement.

In this paper, we propose a design framework for distributed embedded control systems that ensures reliable execution and high quality of control even if some computation nodes fail. When a node fails, the configuration of the underlying distributed system changes and the system must adapt to this new situation by activating tasks at operational nodes. The task mapping as well as schedules and control laws that are customized for the new configuration influence the control quality and must, therefore, be optimized. The number of possible configurations due to faults is exponential in the number of nodes in the system. This design-space complexity leads to unaffordable design time and large memory requirements to store information related to mappings, schedules, and controllers. We demonstrate that it is sufficient to synthesize solutions for a small number of base and minimal configurations to achieve fault tolerance with an inherent minimum level of control quality. We also propose an algorithm to further improve control quality with a priority-based search of the set of configurations and trade-offs between task migration and replication.

Many embedded systems comprise several controllers sharing available resources. It is well known that such resource sharing leads to complex timing behavior that degrades the quality of control, and more importantly, can jeopardize stability in the worst-case, if not properly taken into account during design. Although stability of the control applications is absolutely essential, a design flow driven by the worst-case scenario often leads to poor control quality due to the significant amount of pessimism involved and the fact that the worst-case scenario occurs very rarely. On the other hand, designing the system merely based on control quality, determined by the expected (average-case) behavior, does not guarantee the stability of control applications in the worst-case. Therefore, both control quality and worst-case stability have to be considered during the design process, i.e., period assignment, task scheduling, and control-synthesis. In this paper, we present an integrated approach for designing high-quality embedded control systems, while guaranteeing their stability.

Time-triggered periodic control implementations are over provisioned for many execution scenarios in which the states of the controlled plants are close to equilibrium. To address this inefficient use of computation resources, researchers have proposed self-triggered control approaches in which the control task computes its execution deadline at runtime based on the state and dynamical properties of the controlled plant. The potential advantages of this control approach cannot, however, be achieved without adequate online resource-management policies. This paper addresses scheduling of multiple self-triggered control tasks that execute on a uniprocessor platform, where the optimization objective is to find trade-offs between the control performance and CPU usage of all control tasks. Our experimental results show that efficiency in terms of control performance and reduced CPU usage can be achieved with the heuristic proposed in this paper.

In resource-constrained systems, the interference generated by the concurrent execution of multiple controller tasks leads to extra delay and jitter, which degrade control performance and may even jeopardize the stability of the controlled system. This work presents a general methodology that integrates control issues and real-time schedulability analysis to improve the control performance in embedded systems with time and resource constraints. The performance increase is achieved by properly selecting task periods and deadlines under feasibility constraints.

At runtime, an embedded control system can switch between alternative functional modes. In each mode, the system operates by using a schedule and controllers that exploit the available computation and communication resources to optimize the control performance in the running mode. The number of modes is usually exponential in the number of control loops, which means that all controllers and schedules cannot be produced in affordable design-time and stored in memory. This paper addresses synthesis of multi-mode embedded control systems. Our contribution is a method that trades control quality with optimization time, and that efficiently selects the schedules and controllers to be synthesized and stored in memory.

Many embedded control systems comprise several control loops that are closed over a network of computation nodes. In such systems, complex timing behavior and communication lead to delay and jitter, which both degrade the performance of each control loop and must be considered during the controller synthesis. Also, the control performance should be taken into account during system scheduling. The contribution of this paper is a control-scheduling co-design method that integrates controller design with both static and priority-based scheduling of the tasks and messages, and in which the overall control performance is optimized.

We consider the problem of optimal static period assignment for multiple independent control tasks executing on the same CPU. Previous works have assumed that the control performance can be expressed as a function of the sampling rate only. Arguing that the control delay has a large impact on the control performance, in this work we include the control delay in the cost function. The delay is estimated using an approximate response-time analysis. Assuming linear cost functions for the controllers then allows us to solve the optimal period assignment problem analytically. The performance improvements over previous methods are verified in evaluations on synthetic task sets as well as detailed co- simulations of the controllers, the plants, and the scheduler.

This paper presents a resource management strategy for control tasks that maximizes control performance within the available resources by readjusting the task periods at runtime. A feedback scheduler is used to determine on-line the optimal task periods considering the response over a finite time horizon of the plants controlled by arbitrary linear control laws. We show how this problem can be expressed as an optimization problem, where the objective function relates the sampling periods to the transient responses of the controlled plants, and where restrictions are based on EDF schedulability constraints. For the general case, the solution of the optimization problem is computationally expensive, and thus, an approximate procedure to be executed on-line has been developed. We present simulation results that validate the presented approach

We study the problem of dynamically scheduling a set of state-feedback control tasks controlling a set of linear plants. We consider an on-line non-preemptive scheduling policy that is optimal in the sense that it minimizes a quadratic performance criterion for the overall system. The optimal scheduling decision at each point in time is a function of the states of the controlled plants. To be able to solve the scheduling problem for realistic examples, we use the technique of relaxed dynamic programming to compute suboptimal solutions with error bounds. The approach is compared to earlier approaches in a case study involving simultaneous control of one ball-and-beam process and two DC-servo processes. We also show how the scheduling policy can be modified to allow for background tasks to execute when the need for control is small

This paper presents the incorporation of an auto-tuning real-time garbage collector into a feedback-based on-line scheduling system. The studied feedback scheduler is designed to dynamically adjust the sampling periods of a set of controller tasks in order to maximize the overall control performance. In the suggested approach, the memory management overhead is made explicit and taken into account by the feedback scheduler when scheduling the tasks. It is also described how priorities for memory allocations can be used to control the allocation rates of the application threads in order to optimize the trade-off between memory and CPU time usage. A case study, comparing theoretical analysis and simulated results support the feasibility of the approach.

The real-time implementation of a controller typically introduces artefacts like delay and jitters that have not been considered at the design stage. As a consequence, the system behaves in a non-periodic manner, and the real performance is degraded with respect to the expected response. This paper proposes a hybrid task model to reduce the impact of the scheduling on the control performance. For a large batch of typical plants, we analyze how sensitive the control system is to jitter when the sampling rate is slow or fast compared to the bandwidth of the system

Advances in scheduling theory have given designers of control systems greater flexibility over their choice of timing requirements. Advances in scheduling theory have given designers of control systems greater flexibility over their choice of timing requirements. This could lead to systems becoming more responsive, more flexible and more maintainable. However, experience has shown that engineers find it difficult to exploit these advantages due o the difficulty in determining the "real" timing requirements of systems and therefore the techniques have delivered less benefit than expected. Part of the reason for this is that the models used by engineers when developing systems do not allow for emergent properties such as timing. This paper presents an approach and framework for addressing the problem of identifying an appropriate and valid set of timing requirements and their corresponding control parameters based on a combination of static analysis and simulation.

The paper presents a computational model for real-time control tasks, with the primary goal of simplifying the control and scheduling codesign problem. The model combines time-triggered I/O and inter-task communication with dynamic, reservation-based task scheduling. To facilitate short input-output latencies, a task may be divided into several segments. Jitter is reduced by allowing communication only at the beginning and at the end of a segment. A key property of the model is that both schedulability and control performance of a control task will depend on the reserved utilization factor only. This enables controllers to be treated as scalable real-time components. The model has been implemented in a real-time kernel and validated in a real-time control application.

The paper considers the implementation of digital controllers as real-time tasks in priority-preemptive systems. The performance of a digital feedback control system depends critically on the timing of its sampling and control actions. It is desirable to minimize the computational delay in the controller as well as the sampling jitter and the control jitter. It is shown that by scheduling the two main parts of a control algorithm as separate tasks, the computational delay can often be reduced significantly. A heuristic method for assigning deadlines to the parts is presented. Further modifications are given to reduce the jitter and to facilitate delay compensation. The result is improved control performance under maintained schedulability.