Discovery of Hard-coded External Dependencies in Enterprise Production Environments

Many enterprises perform data-center transformations, consolidations, and migrations to reduce costs and make IT greener. These projects start with discovery of infrastructure and applications and their dependencies. Typically, this is done by network monitoring and middleware configuration analysis. However, certain dependencies may not be detected without code analysis. We designed and implemented the first code-analysis technique for discovering hard-coded dependencies, based on static string analysis. Key novel aspects include the ability to localize the code and associated files in a production enterprise environment, an analysis for identifying functions that can access external resources, and a program-environment analysis (linked to the string analysis) for inferring values originating outside of the program. The approach is sound under reasonable assumptions about the underlying components.

We analyzed 1097 Java EE applications from three enterprise environments. The vast majority had hard-coded dependencies that required our novel analysis techniques. Such applications need special treatment in transformation projects.