Collecting garbage pages in a distributed shared memory with reduced memory and communication overhead

Abstract

We present a novel algorithm for Garbage Collection (GC) in Distributed Shared Memory systems (DSM). Our algorithm reduces the network traffic overhead (and the memory and computation overheads), essentially eliminating all communication when there is no active collection, and minimizing it when the collection process is turned on. Our algorithm works correctly for asynchronous environments where messages may experience arbitrary delays on the way to their destinations. It also tolerates arbitrary duplication of messages and is thus a suitable "add-on" for fault-tolerant communication protocols. It does not suffer from problems such as weight underflow (which arise in reference counting techniques). In addition, when applied in granularity of pages (which is the most relevant in page-based DSM systems), then the memory overhead is not inflated when the average allocation size is small, and the memory reorganization required due to the GC operations is simplified.