Partitioning complex real-time systems for distribution

Abstract

A substantial amount of time and money has been invested in large complex reactive systems. The size of these legacy systems often puts an organization into the position of having too much invested to completely redesign and reimplement the product. Therefore, methods for transitioning these systems to more modern architectures must be synthesized, after the fact, from the existing programming artifacts. Due to the mainframe technology upon which they were based, many of these systems are monolithic and require repartitioning. There is some solace in the fact that many systems have existing software partitionings based on many independent threads of control which become candidates for distributed subsystems. The challenge lies in picking aggregates of these threads which present a reasonable computational load and have loose enough coupling to permit distribution of the final system.<>