Garbage Collection in Concurrent Sets (1806.00834v2)

Abstract: Garbage Collection in concurrent data structures, especially lock-free ones, pose multiple design and consistency challenges. In this instance, we consider the case of concurrent sets. A set is a collection of elements, where the elements are ordered and distinct. These two invariants are always maintained at every point in time. Currently, multiple implementations of concurrent sets already exist. LazyList, Hand-over-hand List and Harris List are some of the well-known implementations. However none of these implementations employ, or are concerned with garbage collection of deleted nodes. Instead each implementation ignores deleted nodes or depends on the garbage collector of the language to handle them. Additionally, Garbage collection in concurrent lists, that use optimistic traversals or that are lock-free, is not trivial. For example, in LazyList and Harris List, they allow a thread to traverse a node or a sequence of nodes after these nodes have already been removed from the list, and hence possibly deleted. If deleted nodes are to be reused, this will potentially lead to the ABA problem. Moreover, some languages like C++ do not have an inbuilt garbage collector. Some constructs like Shared Pointers provide a limited garbage collection facility, but it degrades performance by a large scale. Integrating Shared Pointers into a concurrent code is also not a trivial task. In this paper, we propose a new representation of a concurrent set, GCList, which employs inbuilt garbage collection. We propose a novel garbage collection scheme that implements in-built memory reclamation whereby it reuses deleted nodes from the list. We propose both lock-based and lock-free implementations of GCList. The garbage collection scheme works in parallel with the Set operations.