Each minute of software system crashes and downtime in critical systems is often expensive or dangerous for both humans and the environment. Research reveals an intimidating large effort for maintenance; up to 90% of software cost is derived to maintenance and evolution. The attention on software maintenance has increased due to the safety aspect and its significant economic importance on information technology. Applying a statistical model for planning regular software maintenance action and modification can be misleading because similar software behave and age differently at various operating conditions. Standards that are used for software development and software maintenance do not treat software condition-based or proactive maintenance explicitly. Detecting deviations in the condition and performance of software systems at an early stage should decrease the economic losses.

In this paper, different aspects of critical IT systems have been considered. The major result of this study is; a literature review, classification and analysis of maintenance approaches applied in industry and those in software reinforce that industrial approaches are conceptually comparable with software approaches. Also, investigating the possibility of adapting experiences from industry in the development of a new condition-based maintenance approach regarding software systems is viable. The main conclusions are that static and dynamic properties of software condition can be obtained by using tools and techniques, and if identifying and monitoring software degradation parameters it would be possible to apply a CBM policy on software. Consequently more research effort needs to be accomplished on condition-based software maintenance with a proactive-predictive approach.