Comparative analysis of aperiodic server approaches for real-time garbage collection

Abstract

We analyze the characteristics of the deferrable and sporadic server approaches to real-time garbage collection implemented with single or dual servers. Comparative analysis addresses two problems: the scheduling of a garbage collector with hard real-time mutator tasks, and the worst-case memory requirement. Our analysis shows that it may be impractical to depend on the single-server approach to prepare a flexible, yet feasible schedule for embedded real-time systems. In contrast, dual servers enhance the scheduling flexibility, although it suffers additional overhead of the secondary server and memory space. According to the simulation study, the single-server approach prefers the sporadic server to the deferrable server, since the former has larger capacity than the latter. On the other hand, with dual servers have chosen the deferrable server is the more efficient solution.