DeeDS combines active database functionality with critical timing constraints and integrated system monitoring. Since the reactive database mechanisms, or rule management system, must meet critical deadlines, we must employ methods that make triggering of rules and execution of actions predictable. We will focus on the scheduling issues associated with dynamic scheduling of workloads where the triggered transactions have hard, firm or soft deadlines, and how transient overloads may be resolved by substituting transactions by computationally cheaper ones. The rationale for a loosely coupled general purpose event monitoring facility, that works in tight connection with the scheduler, is presented. For performance and predictability, the scheduler and event monitor are executing on a separate CPU from the rest of the system. Real-time database accesses in DeeDS are made predictable and efficient by employing methods such as main memory resident data, full replication, eventual consistency, and prevention of global deadlocks.
Data science applications often need to deal with data that does not fit into the standard entity-attribute-value model. In this chapter we discuss three of these other types of data. We discuss texts, images and graphs. The importance of social media is one of the reason for the interest on graphs as they are a way to represent social networks and, in general, any type of interaction between people. In this chapter we present examples of tools that can be used to extract information and, thus, analyze these three types of data. In particular, we discuss topic modeling using a hierarchical statistical model as a way to extract relevant topics from texts, image analysis using convolutional neural networks, and measures and visual methods to summarize information from graphs.
Active database rules are problematic to explain, understand, debug, and design irrespective of knowledge about active rule semantics. In order to address this problem, various types of active database tools have been proposed in the literature such as browsers, debuggers, analyzers, and explanation tools.This paper focuses on visualization of event detection for an explanation tool and it presents the first study on what to visualize with respect to event detection at the lowest level (i.e. visualization of event detection for a specific event type).
The test effort required for full test coverage is much higher in an event-triggered than in a time-triggered real-time system. This makes it difficult to attain confidence in the correctness of event-triggered real-time applications by testing, which is a necessary complement to other verification methods. We present a more general upper bound on the test effort of constrained event-triggered real-time systems, assuming multiple resources (a refinement of previous results). The emphasis is on system level testing of application timeliness, assuming that sufficient confidence in its functional correctness has been attained. Covered fault types include incorrect assumptions about temporal attributes of application and execution environment, and synchronization faults. An analysis of the effects that our constraints have on predictability and efficiency shows that the use of designated preemption points is required. A key factor in this approach is the ability to reduce the number of required test cases while maintaining full test coverage.
This study compares seven different methods for handling constraints in input parameter models when using combination strategies to select test cases. Combination strategies are used to select test cases based on input parameter models. An input parameter model is a representation of the input space of the system under test via a set of parameters and values for these parameters. A test case is one specific combination of values for all the parameters. Sometimes the input parameter model may contain parameters that are not independent. Some sub-combinations of values of the dependent parameters may not be valid, i.e., these sub-combinations do not make sense. Combination strategies, in their basic forms, do not take into account any semantic information. Thus, invalid sub-combinations may be included in test cases in the test suite. This paper proposes four new constraint handling methods and compares these with three existing methods in an experiment in which the seven constraint handling methods are used to handle a number of different constraints in different sized input parameter models under three different coverage criteria. All in all, 2568 test suites with a total of 634,263 test cases have been generated within the scope of this experiment.
This paper presents data from a study of the current state of practice of software testing. Test managers from twelve different software organizations were interviewed. The interviews focused on the amount of resources spent on testing, how the testing is conducted, and the knowledge of the personnel in the test organizations. The data indicate that the overall test maturity is low. Test managers are aware of this but have trouble improving. One problem is that the organizations are commercially successful, suggesting that products must already be "good enough". Also, the current lack of structured testing in practice makes it difficult to quantify the current level of maturity and thereby articulate the potential gain from increasing testing maturity to upper management and developers
This report is a survey of monitoring and event detection in distributed fault-tolerant real-time systems, as used in primarily active database systems, for testing and debugging purposes. It contains a brief overview of monitoring in general, with examples of how software systems can be instrumented in a distributed environment, and of the active database area with additional constraints of real-time discussed. The main part is a survey of event monitoring mostly taken from the active database area with additional discussion concerning distribution and fault-tolerance. Similarities between testing and debugging distributed real-time systems are described.
This paper provides a schematic, systematic and structured approach todeveloping Bayesian belief networks to assess risks in contexts dened by activities.The method ameliorates elicitation, specication and validation of expert knowledgeby reusing a schematic structures based on reasoning of risks based on the temporal motivationaltheory. The method is based on earlier work that took a rst signicant steptowards reducing the complexity of development of Bayesian belief networks by clusteringand classifying variables in Bayesian belief networks as well as associating the processwith human deciions making. It may be possible to reduce the role of a facilitiatoror even remove the facilitator altogether by using this method. The method is partiallyvalidated and further work is required on this topic.
We use data-generated models based on data from experiments of an ocean-going vessel to study the effect of optimizing fuel consumption. The optimization is an add-on module to the existing diesel-engine fuel-injection control built by Q-TAGG R&D AB. The work is mainly a validation of knowledge-based models based on a priori knowledge from physics. The results from a simulation-based analysis of the predictive models built on data agree with the results based on knowledge-based models in a companion study. This indicates that the optimization algorithm saves fuel. We also address specific problems of adapting data to existing machine learning methods. It turns out that we can simplify the problem by ignoring the auto-correlative effects in the time series by employing low-pass filters and resampling techniques. Thereby we can use mature and robust classification techniques with less requirements on the data to demonstrate that fuel is saved compared to the full-fledged time series analysis techniques which are harder to use. The trade-off is the accuracy of the result, that is, it is hard to tell exactly how much fuel is saved. In essence, however, this process can be automated due to its simplicity.
In the DeeDS prototype, active database functionality and critical timing constraints are combined with integrated monitoring techniques. In the scope of DeeDS, this paper presents a mathematical model which is used to derive two important design constraints; worst-case minimum delay and maximum frequency of events. This model is based on a dual-processor hybrid-monitoring solution. Furthermore, different interaction styles between the scheduler and the event monitor are evaluated.
In this work, we raise three critical questions that must be investigated to ameliorate composability ofvirtual simulation models and to enable adoption of systematic and stringent real-time techniques toenable more scalable simulation models for virtual and constructive simulation. The real-time techniquesin question enable us to separate between policies and mechanisms and, thus, the simulation engine candecide dynamically how to run the simulation given the existing resources (e.g., processor) and the goalsof the simulation (e.g., sufficient fidelity in terms of timing and accuracy). The three critical questionsare: (i) how to design efficient and effective algorithms for making dynamic simulation model designdecisions during simulation; (ii) how to map simulation entities (e.g., agents) into (real-time) tasks; and(iii) how to enable a divide and conquer approach to validating simulation models.
Temporal correctness is crucial for real-time systems. Few methods exist to test temporal correctness and most methods used in practice are ad-hoc. A problem with testing real-time applications is the response-time dependency on the execution order of concurrent tasks. Execution order in turn depends on execution environment properties such as scheduling protocols, use of mutual exclusive resources as well as the point in time when stimuli is injected. Model based mutation testing has previously been proposed to determine the execution orders that need to be verified to increase confidence in timeliness. An effective way to automatically generate such test cases for dynamic real-time systems is still needed. This paper presents a method using heuristic-driven simulation to generate test cases.
Temporal correctness is crucial for real-time systems. There are few methods to test temporal correctness and most methods used in practice are ad-hoc. A problem with testing real-time applications is the response-time dependency on the execution order of concurrent tasks. Execution orders in turn depends on scheduling protocols, task execution times, and use of mutual exclusive resources apart from the points in time when stimuli is injected. Model-based mutation testing has previously been proposed to determine the execution orders that need to be tested to increase confidence in timeliness. An effective way to automatically generate such test cases for dynamic real-time systems is still needed. This paper presents a method using heuristic-driven simulation for generation of test cases.
A trend over the past years is that simulation systems for training are being connected in simulation networks, allowing the interaction of teams spread in distributed sites. By combining interconnected simulation systems the simulation complexity increases and may affect time-critical simulation tasks in a negative way. As a consequence, the training simulation objectives may not be met. The same problem may occur when performing, for example, mission rehearsal on site, since available computation resources are usually very limited in this scenario, or for a joint fires scenario, where the large and complex functional chain (including intelligence, C2, forward observer, pilots, etc.) may overload existing resources. In this work, the technique of imprecise computation in real-time systems (ICRS) to preserve time-critical simulation tasks is presented. The ICRS technique allows time-critical tasks to produce quicker solutions for approximate results and saves computational resources. This paper discusses the main advantages of theICRS technique by a review of the commonly used optimization concepts built upon imprecise computation field. Thepaper ends with presenting a work-in-progress: an architectural solution for aligning ICRS with the High Level Architecture (HLA), standardized as the IEEE 1516-series.
Using hyperglycemia as an example, we present how Bayesian networks can be utilized for automatic early detection of a person’s possible medical risks based on information provided by un obtrusive sensors in their living environments. The network’s outcome can be used as a basis on which an automated AMI-system decides whether to interact with the person, their caregiver, or any other appropriate party. The networks’ design is established through expert elicitation and validated using a half-automated validation process that allows the medical expert to specify validation rules. To interpret the networks’ results we use an output dictionary which is automatically generated for each individual network and translates the output probability into the different risk classes (e.g.,no risk, risk).