Mining Computing Configurations

点击翻译按钮获取中文摘要

Large scale computing systems, notably IT (Information Technology) systems, consist of elaborate configurations made up of hardware servers connected by networks, operating systems and databases, service software such as Customer Service Management and so on. These configurations evolve with time, and must be constantly monitored and managed to correct faults and to adapt to new demands. It is challenging for analysts to manage and track these changes in order to maximize the reliability and performance of these systems. Increasingly, these configurations are stored in special purpose databases such as Configuration Management Databases (CMDBs). In essence, the CMDB records the present and past architectures of the computing system. It is used to help the analysts control and maintain the system. The research proposed here will use data mining techniques to explore and monitor CMDB-like databases with the long-term goal of assisting analysts to do a better job of running these systems. One of the jobs of the analysts is to plan changes to the system. Such a change requires updating a set of items in the system. It is essential that the plan for the change must include all the items that must be updated to accomplish the whole change. If some items that should be updated are inadvertently overlooked, the attempted change may fail, causing outage, loss of sales, customer dissatisfaction --- even failure of the commercial enterprise. The proposed research will use mining techniques to provide suggestions and advice, based on past patterns of change and activity as recorded in the CMDB to help the analysts avoid problems such as incomplete changes. More generally, the proposed research will search for historic patterns of activity within computing configurations in order to suggests tools and information for decision support to aid analysts.