Performance of a hierarchical distributed garbage collection algorithm in ActorFoundry

Automatic garbage collection is an essential feature so that programs can reclaim resources without the need for manual input. This feature is present in many modern languages and is a common subject of research. However, in parallel and distributed environments, programmer-controlled resource reclamation is highly error-prone. As the scale of programs increase, automatic garbage collection is of paramount importance for efficient and error-free execution. Garbage collection in the context of actor systems is especially difficult because actors are active objects and may not be garbage even if there are no references to it. An additional difficulty is to perform garbage collection on active objects without halting the current computation. This thesis implements one of the proposed algorithms which can solve the problem of garbage collection in distributed actor systems. This study also explores how parameters in this algorithm along with how the topology of an actor system affect the garbage collection. This was implemented on an existing actor framework in order to highlight key factors in the algorithm's performance. The design details and insights gained from the results of these tests are then discussed