Hüseyin Aysan

Maximizing the Fault Tolerance Capability of Fixed Priority Schedules

Real-time systems typically have to satisfy complex requirements, mapped to the task attributes, eventually guaranteed by the underlying scheduler. These systems consist of a mix of hard and soft tasks with varying criticality, as well as associated fault tolerance requirements. Additionally, the relative criticality of tasks could undergo changes during the system evolution. Time redundancy techniques are often preferred in embedded applications and, hence, it is extremely important to devise appropriate methodologies for scheduling real-time tasks under failure assumptions.In this paper, we propose a methodology to provide a priori guarantees in fixed priority scheduling (FPS) such that the system will be able to tolerate one error per every critical task instance. We do so by using integer linear programming (ILP) to derive task attributes that guarantee re-execution of every critical task instance before its deadline, while keeping the associated costs minimized. We illustrate the effectiveness of our approach, in comparison with fault tolerant (FT) adaptations of the well-known rate monotonic (RM) scheduling, by simulations.

VTV - A Voting Strategy for Real-Time Systems

Real-time applications typically have to satisfy high dependability requirements and require fault tolerance in both value and time domains. A widely used approach to ensure fault tolerance in dependable systems is the N-modular redundancy (NMR) which typically uses a majority voting mechanism. However, NMR primarily focuses on producing the correct value, without taking into account the time dimension. In this paper, we propose a new approach, Voting on Time and Value (VTV), applicable to real-time systems, which extends the modular redundancy approach by explicitly considering both value and timing failures, such that correct value is produced at a correct time, under specified assumptions. We illustrate our voting approach by instantiating it in the context of the well-known triple modular redundancy (TMR) approach. Further, we present a generalized version targeting NMR that enables a high degree of customization from the user perspective.

Error Modeling in Dependable Component-Based Systems

Component-based development (CBD) of software, with its successes in enterprise computing, has the promise of being a good development model due to its cost effectiveness and potential for achieving high quality of components by virtue of reuse. However, for systems with dependability concerns, such as real-time systems, a major challenge in using CBD consists of predicting dependability attributes, or providing dependability assertions, based on the individual component properties and architectural aspects. In this paper, we propose a framework which aims to address this challenge. Specifically, we present a revised error classification together with error propagation aspects, and briefly sketch how to compose error models within the context of component-based systems (CBS). The ultimate goal is to perform the analysis on a given CBS, in order to find bottlenecks in achieving dependability requirements and to provide guidelines to the designer on the usage of appropriate error detection and fault tolerance mechanisms.

Probabilistic Schedulability Guarantees for Dependable Real-Time Systems under Error Bursts

The fundamental requirement for the design of effective and efficient fault-tolerance mechanisms in dependable real-time systems is a realistic and applicable model of potential faults, their manifestations and consequences. Fault and error models also need to be evolved based on the characteristics of the operational environments or even based on technological advances. In this paper we propose a probabilistic burst error model in lieu of the commonly used simplistic fault assumptions in the context of processor scheduling. We present a novel schedulability analysis that accounts for the worst case interference caused by error bursts on the response times of tasks scheduled under the fixed priority scheduling (FPS) policy. Further, we describe a methodology for the calculation of probabilistic schedulability guarantees as a weighted sum of the conditional probabilities of schedulability under specified error burst characteristics. Finally, we identify potential sources of pessimism in the worst case response time calculations and discuss potential means for circumventing these issues.

Optimizing the fault tolerance capabilities of distributed real-time systems

Industrial real-time systems typically have to satisfy complex requirements, mapped to the task attributes, eventually guaranteed by a fixed priority scheduler in a distributed environment. These systems consist of a mix of hard and soft tasks with varying criticality, as well as associated fault tolerance requirements. Time redundancy techniques are often preferred in industrial applications and, hence, it is extremely important to devise resource efficient methodologies for scheduling real-time tasks under failure assumptions. In this paper, we propose a methodology to provide a priori guarantees in distributed real-time systems with redundancy requirements. We do so by identifying temporal feasibility windows for all task executions and reexecutions, as well as allocating them on different processing nodes. We then use optimization theory to derive the optimal feasibility windows that maximize the utilization on each node, while avoiding overloads. Finally on each node, we use Integer Linear Programming (ILP) to derive fixed priority task attributes that guarantee the task executions within the derived feasibility windows, while keeping the associated costs minimized.

A Cascading Redundancy Approach for Dependable Real-Time Systems

Dependable real-time systems typically consist of tasks of multiple criticality levels and scheduling them in a faulttolerant manner is a challenging problem. Redundancy in the physical and temporal domains for achieving fault tolerance has been often dealt independently based on the types of errors one needs to tolerate. To our knowledge, there had been no work which tries to integrate fault tolerant scheduling and multiple redundancy mechanisms. In this paper we propose a novel cascading redundancy approach within a generic fault tolerant scheduling framework. The proposed approach is capable of tolerating errors with a wider coverage (with respect to error frequency and error types) than time and space redundancy in isolation, allows tasks with mixed criticality levels, is independent of the scheduling technique and, above all, ensures that every critical task instance can be feasibly replicated in both time and space.

Improving Reliability of Real-Time Systems through Value and Time Voting

Critical systems often use N-modular redundancy to tolerate faults in subsystems. Traditional approaches to N-modular redundancy in distributed, loosely-synchronised, real-time systems handle time and value errors separately: a voter detects value errors, while watchdog-based health monitoring detects timing errors. In prior work, we proposed the integrated Voting on Time and Value (VTV) strategy, which allows both timing and value errors to be detected simultaneously. In this paper, we show how VTV can be harnessed as part of an overall fault tolerance strategy and evaluate its performance using a well-known control application, the Inverted Pendulum. Through extensive simulations, we compare the performance of Inverted Pendulum systems which employs VTV and alternative voting strategies to demonstrate that VTV better tolerates well-recognised faults in this realistically complex control problem.

Probabilistic Schedulability Analysis for Fault Tolerant Tasks under Stochastic Error Occurrences

In dependable real-time systems, provision of schedulability guarantees for task sets under realistic fault and error assumptions is an essential requirement, though complex and tricky to achieve. An important factor to be considered in this context is the random nature of occurrences of faults and errors, which, if addressed in the traditional schedulability analysis by assuming a rigid worst case occurrence scenario, may lead to inaccurate results. In this paper we first propose a stochastic fault and error model which has the capability of modeling error bursts in lieu of the commonly used simplistic error assumptions in processor scheduling. We then present a novel schedulability analysis that accounts for a range of worst case scenarios generated by stochastic error burst occurrences on the response times of tasks scheduled under the fixed priority scheduling (FPS) policy. Finally, we describe a methodology for the calculation of probabilistic schedulability guarantees as a weighted sum of the conditional probabilities of schedulability under specified error burst characteristics.

Schedulability Guarantees for Dependable Distributed Real-Time Systems under Error Bursts

In dependable embedded real-time systems, typically built of computing nodes exchanging messages over reliability-constrained networks, the provision of schedulability guarantees for task and message sets under realistic fault and error assumptions is an essential requirement, though complex and tricky to achieve. An important factor to be considered in this context is the random nature of occurrences of faults and errors, which, if addressed in the traditional schedulability analysis by assuming a rigid worst-case occurrence scenario, may lead to inaccurate results. In this work we propose a framework for end-to-end probabilistic schedulability analysis for real-time tasks exchanging messages over Controller Area Network under stochastic errors.

On voting strategies for loosely synchronized dependable real-time systems

Hard real-time applications typically have to satisfy high dependability requirements in terms of fault tolerance in both the value and the time domains. Loosely synchronized real-time systems, which represent many of the systems that are developed, make any form of voting difficult as each replica may provide different outputs independent of whether there has been an error or not. This can also lead to false positives and false negatives which makes achieving fault tolerance, and hence dependability, difficult. We have earlier proposed a majority voting technique, “Voting on Time and Value” (VTV) that explicitly considers combinations of value and timing errors, targeting loosely-synchronised systems. In this paper, we extend VTV to enable voter parameter tuning to obtain the desired user specified trade-offs between the false positive and false negative rates in the voter outputs. We evaluate the performance of VTV against Compare Majority Voting (CMV), which is a known voting approach applicable in similar contexts, through extensive simulation studies. The results clearly demonstrate that VTV outperforms CMV in all scenarios with lower false negative rates.