#TITLE_ALTERNATIVE#

• Main Author: (NIM 235 04 001), Juwairiah
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/8377
• Institution: Institut Teknologi Bandung

Abstract: <br /> <br /> <br /> <br /> <br /> Many information systems, especially web-based system, are composed of a number of communicating components. These are often structured as distributed systems, with components running on different processors or in different processes. Most of IS often suffered from common problem, that is when failure occur, it is difficult to determine which part of system as the source of failure. Root cause of failure in software system is called fault. Diagnosis manually slow and inconsistent, so it can cause long recovery time dan system availabilty will be low. Therefore, it is necessary to propose another approach to solve this diagnosis problem without detailed knowledge of the system structures and correct behavior with statistic based machine learning approach. Such approach requires less human intervention and therefore is more automatic. <br /> <br /> <br /> <br /> <br /> In this thesis, studied about fault diagnosis in three cases of information systems failure, i.e : bug program, internet services problem, and system performance problem, that use different machine learning approach : Logistic Regression, Decision Tree, and Tree-Augmented-Naive Bayesian Network (TAN) respectivelly. This thesis explore fault diagnosis in three cases of information systems failure, then look for similarity of them, to be foundation in designing a generic framework for automated diagnosis of distributed information systems failure. <br /> <br /> <br /> <br /> <br /> From exploration of fault diagnosis in three cases of diagnosis IS failure, can be inferred that there are 4 components that have basic function. The four basic components are Sensor (collect the data that associated with failure), LogDB (repository of information that is taken from log file of application and will be used as training set), machine learning with statistical analyzer (to classify and analyze data), and conclusion (give conclusion abaout root cause of the problem from computation of statistical analyzer). <br /> <br /> <br /> <br /> <br /> Designing consist of two design : designing phase of fault diagnosis and designing of classes for framework. Designing phase of fault diagnosis is to design steps or phase that done as frame of thinking in do fault diagnosis. Then do designing of classes for framework for each case, and make a generic framework to diagnosis IS failure based on similarity of fault diagnosis in three cases of information systems failure. <br /> <br />