Predictive Algorithms and Technologies for Availability Enhancement

摘要

Predicting the future has fascinated people throughout civilizations but until the 20th century it has been more of a magic than science. Ability to predict the future has a significant impact on wide spectrum of applications ranging from business, communication systems and politics to health monitoring and enviromental protection. It will also become essential in dependability enhancement, security and critical infrastructure protection. With development of machine learning, pattern recognition, data mining and computer technology we seem to be better equipped to tackle the prediction problem with a more science-based, goal-oriented approach. In this tutorial, we focus on predictive algorithms and technologies which may have a major impact on computer systems availability, security and performance. We first motivate our approach, present the best practice guide for using prediction algorithms and models, then briefly survey short-term prediction techniques, introduce two concrete prediction methods and prediction quality measures, and then demonstrate how the availability of software and hardware systems can be increased by preventive measures which are triggered by short-term failure prediction mechanisms.