Dariusz Caban

Estimation of repairable system availability within fixed time horizon

The paper presents an efficient approach to estimating the availability of a repairable system observed within a fixed time horizon, which is regarded as a random variable over range [0,1]. A Beta distribution based approximation of this characteristic is proposed. The presented results are derived from Monte Carlo simulation and statistical analysis of the simulation results. Numerical examples are also provided to illustrate the practical impact of this approach.

Introduction to the Dependability Modeling of Computer Systems

Computer systems and networks are considered as a union of all resources, i.e. hardware, software and people (users, administrators and managers), essential for the realization of predicted tasks. The system dependability is defined as a generalization of performability and reliability, combining the notions of both these terms. The functional-reliability models are based on the observation that only a subset of the system resources is involved in a task execution and only inefficiencies of these resources may influence the correctness of task realization. The systems are working in a real environment, which is often hostile: it may be a source of threats, such as security intrusions, faulty or modified software and human errors. The set of analyzed events is a sum of hardware failures and malfunctions, software faults, human errors, intrusions (intended and addressed threats), and viruses (unaddressed threats that are broadcast in the system and in its environment)

Dependability Analysis of SOA Systems

Service oriented architecture solutions are becoming increasingly popular in business processes. One of challenges this technology has to face is providing appropriate dependability levels and means for their calculation. There is a growing need for SOA solutions assessment before deployment. In this paper we try to develop a model of a business level service provider and define few metrics. The described model has been implanted into SSFNet simulator and some testing has been done. We present the results of our experiments.

Dependability analysis of reconfigurable information systems

Complex network information systems are usually responsible for realizing client requests of business level services. Often these services are critical for the company and it is very important to ensure that appropriate reconfiguration will be applied in case of system malfunction. Due to the complexity of systems, huge amount of dependencies and impacting factors, the analytical approach may be too limiting to evaluate the reconfiguration policy. The paper presents a method of analyzing new applicable configurations of information systems based on simulation. As a result, the reachability graph of configurations is constructed. Each transition in the graph defines the most adequate reconfiguration (compliant with user criteria) in case of a system failure.

FPGA Implementation of Positional Filters

The paper reports on some experiments with implementing positional digital image filters using field programmable devices. It demonstrates that a single field programmable device may be used to build such a filter. By using extensive pipelining in the design, the filter can achieve performance of 50 million pixels per second (using Xilinx XC4000E devices) and over 120 MHz (in case of Spartan-3 devices). These results were obtained using automatic synthesis from VHDL descriptions, avoiding any direct manipulation in the design.

Dependability Analysis of Information Systems with Hierarchical Reconfiguration of Services

The paper presents an approach to dependability analysis of a class of networked information systems based on distributed services. The system consists of groups of servers that react autonomously to detected events, component failures and security incidents. These local reactions affect the availability of the services, provided by the system, then restored by reconfiguration. The dependability analysis is based on the construction of a reconfiguration graph, which is transformed to a state-transition dependability model.

Improvement of Dependability of Complex Web Based Systems by Service Reconfiguration

Web based information systems are exposed to various dependability issues during their lifetime (originating in the hardware, in the software or stemming from security vulnerabilities). We present an approach to use reconfiguration to circumvent these issues. The presentation is focused on the potential threads, measuring the dependability risks, constructing of optimal reconfiguration strategies and assessing their impact on the over-all dependability. The proposed technique involves construction of the reconfiguration graph, assessment of the steady state probabilities of web system dependability states and choosing the optimal strategy from among the admissible ones.

Prediction of the Performance of Web Based Systems

Complex Web based information systems are organized as a set of component services, communicating using the client-server paradigm. The performance prediction of such systems is complicated by the fact that the service components are strongly inter-dependent. To overcome this issue, it is proposed to use simulation techniques. Extensions to the available network simulation tools are proposed to support this. The authors present the results of multiple experiments with web-based systems, which were conducted to develop a model of client-server interactions adequately describing the relationship between the server response time and resource utilization. This model was implemented in the simulation tools and its accuracy verified against a testbed system configuration.

Reliability Analysis of Discrete Transportation Systems Using Critical States

The paper presents a resource constrained model of a discrete transportation system that can be used to simulate its operation in presence of faults. The simulation results are used to determine the initial level of resources that ensures seamless operation of the system. The simulator is also used to assess the conditional probability of system failure after reaching a specific set of reliability states. This is used to determine the set of critical states, i.e. the states when the system is still operational but the probability of failure in near future is unacceptably high. These states can be used as an indicator that the system is degrading dangerously.

Service Availability Model to Support Reconfiguration

Web based information systems are exposed to various dependability issues during their lifetime. Reconfiguration is then used by the administration to ensure continuity of service. Such forced reconfigurations can cause unforeseen side-effects, such as server overloading. To prevent this, it is proposed to use simulation techniques to analyze the reconfigurations and construct a safe reconfiguration strategy. Extensions to the available network simulation tools are proposed to support this. The authors present the results of multiple experiments with web-based systems, which were conducted to develop a model of client-server interactions that would adequately describe the relationship between the server response time and resource utilization. This model was implemented in the simulation tools and its accuracy verified against a testbed system configuration.