Empirical results on long-lived renaming algorithms

The Long-lived Renaming problem is an important subject in Distributed Algorithms. The Renaming problem consists in providing processes with new names from a hopefully smaller name space. In the long-lived version, only a fraction of the processes request a new name and later on abandon it and this name could be acquired by another processor. A bound is assumed on the number of processors holding simultaneously a name from the smaller name space. Only recently wait-free fast algorithms for long-lived renaming on asynchronous and shared memory distributed systems have been found and analyzed, and they had not been actually implemented. We describe some empirical considerations complementing the formal static analysis, and assess how useful each of them could be in a real application, as a result of the study of their implementations on a running asynchronous distributed system.