東華大學圖書館 |

AI-Enabled Log Analysis for Improving It System Dependability /

Record Type:	Electronic resources : Monograph/item
Title/Author:	AI-Enabled Log Analysis for Improving It System Dependability // Jasmin Bogatinovski.
Author:	Bogatinovski, Jasmin,
Description:	1 electronic resource (195 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-11, Section: B.
Contained By:	Dissertations Abstracts International85-11B.
Subject:	Software. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31077197
ISBN:	9798382644233

AI-Enabled Log Analysis for Improving It System Dependability /
Bogatinovski, Jasmin,

AI-Enabled Log Analysis for Improving It System Dependability /Jasmin Bogatinovski. - 1 electronic resource (195 pages)

Source: Dissertations Abstracts International, Volume: 85-11, Section: B.

Modern IT systems play an indispensable role in industrial infrastructure and affect human society, as billions of users and devices constantly compute, exchange and store data. Their characteristics, such as large complexity, fast evolution, and geo-distributed development, among others, challenge the availability and the correctness of service offerings while increasing failure proneness. Failure to deliver the correct service can have severe implications. This is particularly the case for critical systems in medicine, transportation, or energy, leading to hazardous effects. The increased complexity surpasses the developers' and operators' capabilities for timely issue resolution increasing the chance of frequent failure impact.To support the system development and operation, as means to ensure the provisioning of correct service that can justifiably be trusted (system dependability), automation of different tasks is needed. One important aspect of automation is the IT system's capability to externalise the system state via monitoring data such as system logs. These data are used by intelligent methods that can learn to discern frequent normal and anomalous patterns from the data. Therefore, intelligent methods can automate parts of the development and operational processes, e.g., by generating alerts about potential issues. In this context, Artificial Intelligence for IT operations (AIOps) emerged as a research area concerned with using the system (e.g., source code) and monitoring data (e.g., system logs) and methods from artificial intelligence (AI), big data, machine learning and data mining to support the automation of IT operational activities.This thesis introduces AI-enabled methods that address different AIOps tasks during system development and operation. The methods focus on the logging process and system logs as an intrinsic data source for the IT systems. From a system development perspective, the main contributions reside in formalizing and addressing the problem of log instruction quality, as logs with sufficient quality are a precondition for successfully tackling downstream log-related tasks. (1) The thesis proposes a deep learning-based method to automatically evaluate the quality of log instructions from the system's source code. From a system operation perspective, the thesis contributes by proposing novel methods for log analysis, specifically, log-based anomaly detection. The log-based anomaly detection methods learn anomaly-related log properties that improve the (2) sentiment and (3) sequential log representations. This category of methods studies how leveraging the individual log properties impacts anomaly detection and classification in modern IT systems. The extensive evaluations with data from open-source, production systems, and testbeds show the usefulness of the proposed methods in addressing the challenges of modern IT systems while demonstrating desirable practical properties. The proposed methods and results were published in peer-reviewed international conferences, while parts were patented at the European Patent Office.

English

ISBN: 9798382644233Subjects--Topical Terms:

619355
Software.

AI-Enabled Log Analysis for Improving It System Dependability /
LDR:08352nmm a22004453i 4500 001 2437666
005 20251124054954.5
006 m o d
007 cr|nu||||||||
008 261011s2023 miu||||||m |||||||eng d
020 $a 9798382644233
035 $a (MiAaPQD)AAI31077197
035 $a (MiAaPQD)TechBerlin_20231130318305
035 $a AAI31077197
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Bogatinovski, Jasmin, $e author. $3 3828732
245 1 0 $a AI-Enabled Log Analysis for Improving It System Dependability / $c Jasmin Bogatinovski.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2023
300 $a 1 electronic resource (195 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-11, Section: B.
500 $a Advisors: Kao, Odej; Natella, Roberto; Bermbach, David; Madjarov, Gjorgji.
502 $b Ph.D. $c Technische Universitaet Berlin (Germany) $d 2023.
520 $a Modern IT systems play an indispensable role in industrial infrastructure and affect human society, as billions of users and devices constantly compute, exchange and store data. Their characteristics, such as large complexity, fast evolution, and geo-distributed development, among others, challenge the availability and the correctness of service offerings while increasing failure proneness. Failure to deliver the correct service can have severe implications. This is particularly the case for critical systems in medicine, transportation, or energy, leading to hazardous effects. The increased complexity surpasses the developers' and operators' capabilities for timely issue resolution increasing the chance of frequent failure impact.To support the system development and operation, as means to ensure the provisioning of correct service that can justifiably be trusted (system dependability), automation of different tasks is needed. One important aspect of automation is the IT system's capability to externalise the system state via monitoring data such as system logs. These data are used by intelligent methods that can learn to discern frequent normal and anomalous patterns from the data. Therefore, intelligent methods can automate parts of the development and operational processes, e.g., by generating alerts about potential issues. In this context, Artificial Intelligence for IT operations (AIOps) emerged as a research area concerned with using the system (e.g., source code) and monitoring data (e.g., system logs) and methods from artificial intelligence (AI), big data, machine learning and data mining to support the automation of IT operational activities.This thesis introduces AI-enabled methods that address different AIOps tasks during system development and operation. The methods focus on the logging process and system logs as an intrinsic data source for the IT systems. From a system development perspective, the main contributions reside in formalizing and addressing the problem of log instruction quality, as logs with sufficient quality are a precondition for successfully tackling downstream log-related tasks. (1) The thesis proposes a deep learning-based method to automatically evaluate the quality of log instructions from the system's source code. From a system operation perspective, the thesis contributes by proposing novel methods for log analysis, specifically, log-based anomaly detection. The log-based anomaly detection methods learn anomaly-related log properties that improve the (2) sentiment and (3) sequential log representations. This category of methods studies how leveraging the individual log properties impacts anomaly detection and classification in modern IT systems. The extensive evaluations with data from open-source, production systems, and testbeds show the usefulness of the proposed methods in addressing the challenges of modern IT systems while demonstrating desirable practical properties. The proposed methods and results were published in peer-reviewed international conferences, while parts were patented at the European Patent Office.
520 $a Moderne IT-Systeme spielen eine unverzichtbare Rolle in industriellen Infrastrukturen und beeinflussen menschliche Gesellschaften, da Milliarden von Nutzern und Geraten standig Daten berechnen, austauschen und speichern. Diese Systeme weisen Merkmale wie hohe Komplexitat sowie schnelle und geografisch verteilte Entwicklung auf, womit sie eine Herausforderung fur die Verfugbarkeit und Korrektheit von Dienstangeboten darstellen und die Fehleranfalligkeit erhohen. Wird ein Dienst nicht korrekt erbracht, kann dies schwerwiegende Folgen haben. Dies gilt insbesondere fur kritische Systeme in der Medizin, im Transportwesen oder im Energiesektor, was zu gefahrlichen Auswirkungen fuhren kann. Die zunehmende Komplexitat ubersteigt die Moglichkeiten der Entwickler und Betreiber zur rechtzeitigen Problemlosung und erhoht die Wahrscheinlichkeit haufiger Ausfalle.Zur Unterstutzung der Systementwicklung, des Systembetriebs und als Mittel zur Sicherstellung der Bereitstellung eines korrekten Dienstes, dem man berechtigterweise vertrauen kann (Systemzuverlassigkeit), ist die Automatisierung verschiedener Aufgaben erforderlich. Ein wichtiger Aspekt der Automatisierung ist die Fahigkeit des IT-Systems, den Systemzustand uber Uberwachungsdaten wie System-Logs zu externalisieren. Diese Daten werden von intelligenten Methoden verwendet, die lernen konnen, normale und anomale Muster aus den Daten zu erkennen. Intelligente Methoden konnen daher Teile der Entwicklungs- und Betriebsprozesse automatisieren, indem sie z. B. Warnungen oder Korrekturmasnahmen zu potenziellen Problemen erzeugen. In diesem Zusammenhang hat sich Kunstliche Intelligenz fur den IT-Betrieb (AIOps) als Forschungsgebiet herauskristallisiert, das sich mit der Nutzung von System- (z. B. Quellcode) und Uberwachungsdaten (z. B. System-Logs) und Methoden aus den Bereichen KI, Big Data, maschinelles Lernen und Data Mining beschaftigt, um die Automatisierung von IT-Betriebsaktivitaten zu unterstutzen.In dieser Arbeit werden KI-gestutzte Methoden vorgestellt, die verschiedene AIOps-Aufgaben wahrend der Systementwicklung und des Betriebs adressieren. Die Methoden konzentrieren sich auf den Log-Prozess und die System-Logs als intrinsische Datenquelle fur die IT-Systeme. Aus der Perspektive der Systementwicklung liegen die Hauptbeitrage in der Formalisierung und Behandlung des Problems der Qualitat von Log-Instruktionen, da Logs mit ausreichender Qualitat eine Voraussetzung fur die erfolgreiche Bewaltigung nachgelagerter logbezogener Aufgaben sind. (1) Die Arbeit schlagt eine auf Deep Learning basierende Methode zur automatischen Bewertung der Qualitat von Log-Anweisungen aus dem Quellcode des Systems vor. Aus Sicht des Systembetriebs leistet die Arbeit einen Beitrag, indem sie neuartige Methoden zur Log-Analyse vorschlagt, insbesondere zur log-basierten Anomalie-Erkennung. Die log-basierten Methoden zur Erkennung von Anomalien lernen anomalitatsbezogene Log-Eigenschaften, die die (2) Sentiment-und (3) sequentielle Darstellung von Logs verbessern. Diese Kategorie von Methoden untersucht, wie sich die Nutzung der individuellen Log-Eigenschaften auf die Anomalieerkennung und -klassifizierung in modernen IT-Systemen auswirkt. Die umfangreichen Auswertungen mit Daten aus Open-Source- und Produktionssystemen sowie Testumgebungen unterstreichen die Nutzlichkeit der vorgeschlagenen Methoden bei der Bewaltigung der Herausforderungen moderner IT-Systeme und demonstrieren gleichzeitig die wunschenswerten praktischen Eigenschaften. Die vorgeschlagenen Methoden und Ergebnisse wurden von Fachleuten uberpruft und auf internationalen Konferenzen in Form von Fachbeitragen veroffentlicht, sowie in Teilen beim Europaischen Patentamt patentiert.
546 $a English
590 $a School code: 1119
650 4 $a Software. $2 gtt. $3 619355
650 4 $a Failure. $3 3561225
650 4 $a Deep learning. $3 3554982
650 4 $a Linguistics. $3 524476
650 4 $a Semantics. $3 520060
650 4 $a Computer science. $3 523869
650 4 $a Logic. $3 529544
690 $a 0290
690 $a 0800
690 $a 0984
690 $a 0395
710 2 $a Technische Universitaet Berlin (Germany). $e degree granting institution. $3 3819406
720 1 $a Kao, Odej $e degree supervisor.
720 1 $a Natella, Roberto $e degree supervisor.
720 1 $a Bermbach, David $e degree supervisor.
720 1 $a Madjarov, Gjorgji $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 85-11B.
790 $a 1119
791 $a Ph.D.
792 $a 2023
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31077197