Brief announcement: A concurrent partial snapshot algorithm for large-scale and dynamic distributed systems

Abstract

We propose a concurrent partial snapshot algorithm (CSS algorithm) to extend a previously proposed sub-snapshot algorithm (SSS algorithm) [2] by introducing a method of merging multiple snapshots that are concurrently initiated by different nodes. In earlier work [5,6], efficient merging algorithms have already been introduced for CL algorithm. On the other hand, the main issue of our merging algorithm is to cope with dynamic situations based on SSS algorithm. Since the SSS algorithm is an extension of Chandy-Lamport snapshot algorithm (CL algorithm) [1], it allows large-scale and dynamic situations in snapshots. A dynamic situation means that nodes can join and leave freely during the execution of a snapshot algorithm. A snapshot algorithm for the dynamic situation has also been proposed [4]; however in this algorithm, nodes must stop sending application messages during its execution of the snapshot algorithm. Moreover, for concurrent snapshots, it has to cancel a portion of snapshot algorithms. Our algorithm has successfully removed these restrictions. © 2011 Springer-Verlag.