Advances in Distributed Systems pp 104-126
Recent Advances in Distributed Garbage Collection
We have described a detector of distributed cycles of garbage. Our cycle detector presents some interesting properties for large-scale systems: asynchrony between participating spaces, optional participation, tolerance to communication faults and space crashes, low resource requirements, and ease of implementation (no modifications to local objects, only a few to the local garbage collector). Moreover, this algorithm has already been implemented in a distributed system, and the implementation details can be found in le Fessant .
We are now working on a new version of this cycle detector using both propagation of marks and back-tracing. This new algorithm will have the same properties as the one described here, plus the ability to detect cycles spanning any spaces in the whole system. This new algorithm is currently being implemented for our mobile agents platform .
Unable to display preview. Download preview PDF.