Browsing by Author "Zwietasch, Tim"

Now showing 1 - 2 of 2

Open Access
Detecting anomalies in system log files using machine learning techniques
(2014) Zwietasch, Tim
Log files, which are produced in almost all larger computer systems today, contain highly valuable information about the health and behavior of the system and thus they are consulted very often in order to analyze behavioral aspects of the system. Because of the very high number of log entries produced in some systems, it is however extremely difficult to find relevant information in these files. Computer-based log analysis techniques are therefore indispensable for the process of finding relevant data in log files. However, a big problem in finding important events in log files is, that one single event without any context does not always provide enough information to detect the cause of the error, nor enough information to be detected by simple algorithms like the search with regular expressions. In this work, three different data representations for textual information are developed and evaluated, which focus on the contextual relationship between the data in the input. A new position-based anomaly detection algorithm is implemented and compared to various existing algorithms based on the three new representations. The algorithms are executed on a semantically filtered set of a labeled BlueGene/L log file and evaluated by analyzing the correlation between the labels contained in the log file and the anomalous events created by the algorithms. The results show, that the developed anomaly detection algorithm generates the most correlating set of anomalies by using one of the three representations.
Open Access
Online failure prediction for microservice architectures
(2017) Zwietasch, Tim
In many modern software architectures, failure avoidance strategies are already an integral part of the system since they provide many ways to contribute to software resilience. Failures are the cause of system downtimes and latencies and often, they can not be completely prevented. In contrast to fully virtualized servers on which applications are run, some microservice architectures allow microservies to operate natively on the underlying OS and they might therefore interact with each other on a much higher level. Microservices that are deployed on the same node may also affect each other much more than VM’s, for example by putting a high workload on the underlying host. Traditional VM’s run their own Operating system, often in an isolated memory region and a predetiermined, mostly static CPU share whereas microservices are able to cooperate by sharing the same IP-address and other resources. The goal of this thesis is to show how and to which degree microservices can affect each other when they are being executed on the same host and to discuss the effects that these side effects have on failure predictors. For this, a number of simulations are run on a selected containerized application that demonstrate the container-induced side effects. Certain metrics like the CPU-usage of the containers will be evaluated for each scenario and online failure prediction methods are implemented that try to forecast failures based on these metrics. The results show, that independent microservices can affect each other in various ways, for example, by over-utilizing the CPU resources of the host on which they are deployed on. This effect makes failure prediction with monolithic approaches that do not consider the architecture of the host very difficult. This thesis shows and discusses various scenarios in which hierarchical failure prediction methods show significantly better results than monolithic aproaches when such a side-effect is introduced into the system.