Andreas Pietzowski

Prediction of indoor movements using bayesian networks

This paper investigates the efficiency of in-door next location prediction by comparing several prediction methods. The scenario concerns people in an office building visiting offices in a regular fashion over some period of time. We model the scenario by a dynamic Bayesian network and evaluate accuracy of next room prediction and of duration of stay, training and retraining performance, as well as memory and performance requirements of a Bayesian network predictor. The results are compared with further context predictor approaches – a state predictor and a multi-layer perceptron predictor using exactly the same evaluation set-up and benchmarks. The publicly available Augsburg Indoor Location Tracking Benchmarks are applied as predictor loads. Our results show that the Bayesian network predictor reaches a next location prediction accuracy of up to 90% and a duration prediction accuracy of up to 87% with variations depending on the person and specific predictor set-up. The Bayesian network predictor performs in the same accuracy range as the neural network and the state predictor.

A new adaptive accrual failure detector for dependable distributed systems

The detection of failures in distributed environments is a crucial part for developing dependable, robust, and self-healing systems. The contribution of this paper is a new failure detection algorithm that can be described as an adaptive accrual algorithm coupled with features to increase flexiblity and decrease computation costs. Furthermore our evaluation results show a very good detection quality in the case of message losses.

A Lazy Monitoring Approach for Heartbeat-Style Failure Detectors

Failure detectors are a fundamental part of safe fault-tolerant distributed systems. Many failure detectors use heartbeats to draw conclusions about the state of nodes within a distributed environment. The contribution of this paper is an approach whose benefits are twofold. On the one hand it reduces the network overhead produced by heartbeat-style failure detectors. On the other hand it improves the quality of these failure detectors by providing them with richer information about the current network condition. We call this approach lazy monitoring since the active sending of heartbeats is avoided if possible. As it is independent of the actual failure detection algorithm it can be used in many domains. For evaluation purposes we applied our approach to the Smart Doorplate Project. In this testbed the proposed technique reduced the traffic to 1.2% while providing much more information about the environment to the failure detectors.

Adaptive Self-optimization in Distributed Dynamic Environments

Grid and ubiquitous computing systems generally consist of a large number of networked nodes with applications implemented as distributed services or processes, respectively. A crucial point is the distribution of the services to balance the load within the system during runtime. In a former work we developed a self-optimization mechanism which shows outstanding performance in static environments where the services do not change their resource consumptions (e.g. CPU, memory, communication bandwidth). In this paper we present simulation results for the self-optimization within dynamic environments where the services change their load during runtime.

Self-Configuration and Self-Healing in AUTOSAR

State-of-the-art automobiles contain up to 70 microcontrollers. About 35% of all failures are caused by the electronic systems, either hardware or software. Furthermore the stockkeeping of spare parts is a cost intensive issue. On the other hand, microcontrollers evolve so fast that the used microcontrollers cannot be replaced by newer controller generations without any side effects. AUTOSAR [4,5] is a first step to overcome these limitations in the design and maintenance of cars. Based on the AUTOSAR architecture we propose an organic middleware, which adds self-configuration and selfhealing capabilities. We implemented a simulator to evaluate the self-configuration and self-healing and we present the evaluation results for different network sizes and a failure rate of up to 60%. Our results show an excellent performance for the self-configuration in terms of the amount of messages and that the self-healing is able to recover from failures as long as enough resources are available.

Using Automated Planning for Trusted Self-organising Organic Computing Systems

The increasing complexity of computer-based technical systems require new ways to control them. The initiatives Organic Computingand Autonomic Computingaddress exactly this issue. They demand future computer systems to adapt dynamically and autonomously to their environment. In this paper we propose a new approach based on automated planning to realise self-organising capabilities for complex distributed computing systems. The user/administrator only defines objectives describing the conditions which should hold in the system, whereas the system itself is responsible for meeting them using a planning engine. As many planning algorithms are known to be sound and complete, formal guarantees can be given. Thus we aim at building trusted self-organising distributed computer system which are suitable to control real technical systems. Our approach is demonstrated and evaluated on the basis of a simulated production cell with robots and carts. We propose and evaluate two optimisations.

Variations and evaluations of an adaptive accrual failure detector to enable self-healing properties in distributed systems

The initiatives Organic Computing and Autonomic Computing introduced challenging visions for future computer systems. They address the growing complexity of these systems that demands for new ways to control them. Future systems should be able to adapt dynamically to the current conditions of their environment. They should be characterised by so-called self-x properties like self-configuring, self-healing, self-optimising, self-protecting, and context-aware. For the incorporation of self-healing capabilities into distributed systems the detection of failures is a crucial part. Recently we proposed a new failure detector that can be described as an adaptive accrual algorithm. It has been designed for flexible generic usability as a basis to realise self-healing of distributed systems. This paper introduces variations of the proposed basic algorithm to improve its performance and provides an evaluation of all algorithms using message delay and loss models of the internet.

A bio-inspired approach for self-protecting an organic middleware with artificial antibodies

Our human body is well protected by antibodies from our biological immune system. This protection system matured over millions of years and has proven its functionality. In our research we are going to transfer some techniques of a biological immune system to a computer based environment. Our goal is to design a self-protecting middleware which is not vulnerable to malicious events. First off this paper proposes an artificial immune system and evaluates optimal parameter settings. This shows the correlation between the size of a system and the length of the receptors used within antibodies for an efficient detection. Our tests showed that the recognition rate of unknown malicious objects can reach up to 99%. Further on we describe the integration of the immune system into our organic middleware OCμ and afterwards we propose techniques to minimize the memory space needed for storing the antibodies and to speedup the time needed for detecting malicious messages. We obtained a space minimization by 30% and gained a speedup of 30 with execution time optimization.

Autonomous and scalable failure detection in distributed systems

The growing complexity of distributed systems makes it more and more difficult to manage them. Therefore, it is necessary that such systems will be able to adapt autonomously to their environment. They should be characterised by so-called self-x properties such as self-configuration or self-healing. The autonomous detection of failures in distributed environments is a crucial part for developing self-healing systems. In this paper, we introduce algorithms to form monitoring relations and propose to utilise these for a scalable autonomous failure detection. The evaluation of the developed algorithms indicates that they are suitable for complex, large scale and distributed systems.

A distributed self-healing data store

Due to the huge amount of integrated devices and sensors in everyday objects ubiquitous systems are in vicinity and will be deployed in large scales in the near future.We expect these system to be unreliable as nodes may crash or vanish from time to time. Therefore a reliable data store is needed to offer application developers a secure place to store the data of the services. The data store itself is subject to the same unreliable infrastructure thus it must expose self-healing capabilities to overcome data loss due to node failures. In this paper we propose a distributed self-healing data store for ubiquitous systems that guarantees the availability of the stored data even if there is a node failure every 36 seconds in a system consisting of 100 nodes. We also monitor the availability of the nodes to improve the way the data of the data store is distributed in the system.