Maintenance and transformation of large program systems require means for com- prehension of their data dependences. We consider facilities for data flow visualization implemented in the RescueWare® system — a workbench for modernization of legacy systems. One of the key points in the process is the user-guided identification and extrac- tion of business rules, which could be performed successfully only if a user can efficiently trace the data flow. A typical legacy system consists of a large number of comparatively small modules that extensively exchange data either directly or through files, screens, and databases, and hence the analysis cannot be restricted to a single module. On the other hand, it is practically impossible and useless to show the complete data flow graph of the system. We present some techniques of taming the data flow complexity.