IBM carried out an analysis of error categories across a range of its own and third-party products.  Help desk personnel and product experts carried out the analysis and found fewer than 24 primary exception categories. This low number enables the creation of standard log file formats and associated entries. The challenge for software developers is to use this standard format for describing events. What do we do about legacy applications that generate ad hoc log files? We can't just change all existing applications so that their log files conform to the CBE guidelines. Figure 2 illustrates the IBM approach, in which each application, server, database, and so on generates its own domain-specific legacy log file.
Figure 2 Adapting legacy log file data for autonomic computing.
The adapters illustrated in Figure 2 are middleware applications used to translate the legacy log files into CBE-format files. These can then be analyzed by an autonomic manager application, which latter tries to produce an overall analysis of the transformed log file data. Policies and symptom services can then be looked up by the autonomic manager and used as a basis for feeding back corrective data to the applications. In this context, the elements on the left side of Figure 2 are "managed resources."
The system design in Figure 2 closes the loop and creates an automatic feedback system. Human input is removed. While this is not quite the same as the self-aware "Skynet" from the Terminator movies, it has its own problems: When should human intervention occur? Will the system as shown be stable? These are important issues.