Garbage collection of a distributed heap
Rivka Ladin, Barbara H. Liskov · 2003
A practical, fault-tolerant method for reclaiming inaccessible objects in a distributed heap is presented. The algorithm is general and does not require homogeneous components. It reclaims inaccessible objects in a timely fashion, including those that reside on inaccessible cycles. It allows each computer that contains parts of the heap to garbage collect independently according to its storage requirements, using whatever algorithm it chooses. A highly available service is used to store information about the intercomputer references. The computers containing parts of the heap communicate with the central service only periodically. By using the service the overhead at each node is minimized.>