Research Of IT System Self-healing Framework Based On Process Model

One of the grandest challenges faced by IT industry is the system management complexity problem. Frequent system malfunction, more demanding skills on the support staff, and increasing operation management cost are the salient symptoms of this problem. Software Engineering didn't pay enough attention to the research of software system in operation phase. IT operation management process standardization can not reduce system management complexity as well. Self-management, as the key toward autonomic computing, is the main research area in dealing with this grandest challenge. "Self-healing" is one of the most improtant function in self-management. It aims to improve the capability of a software system in dealing with bugs and adapting to changes in its environment. It can eliminate salient symptoms of IT system management complexity problem. Self-healing forms an area of research that is intuitively appealing and is garnering increased attention.This paper proposes an IT system self-healing framework based on process model. Process model supports self-healing by implementing process improvement loop. The loop is composed of four phrases including define, measure, correlate, and improve. These phases implemente the logical abstract, metrics program, correlate model, and improvement method respectively. As an open technique framework, process model improve system through identify and control the vital key factors which can cause exception. Thus it can improve the system reliability and service quality while reduce the need of human intervention. Research on process model extendes the Software Engineering method who is focusing on software development. This paper implemented two process models which provide self-healing capability toward different types of exception. One is Extended SPC(ESPC) model which deals with random performance exception. The other is Function Point Slicing Model which manages system functional exception.ESPC model is introduced to prevent or self-heal performance exception. ESPC enhances the metrics program of Statistical Process Control(SPC). It can dynamically adjust the threshold of performance exception in accordance with system load. Thus it can classify performance data dynamically. In order to measure process performance from user experience perspective, ESPC proposed a metrics called Process Performance Index(PPI). PPI can statistically calculate the percentage of process performance which meets user's expectation. Different processes can compare their performance based on PPI. Correlation model in ESPC can identify the vital few factors which cause performance exception, based on the classified performance data and the synchronous data of various factors. Two performance exception self-healing case studies, in the domain of network proxy service and combinatorial optimization algorithm respectively, are provided. These two case studies show that ESPC can measure and analyze performance from user's perspective, and can discover domain knowledge to support self-healing improvement.