Decomposable algorithms for computing minimum spanning tree

In the emerging networked environments computational tasks are encountering situations in which the datasets relevant for a computation exist in a number of geographically distributed databases, connected by wide-area communication networks. A common constraint in such situations of distributed data is that the databases cannot be moved to other network sites due to security, size, privacy or data-ownership considerations. For these situations we need algorithms that can decompose themselves at run-time to suit the distribution of data. In this paper we present two such self-decomposing algorithms for computing minimum spanning tree for a graph whose components are stored across a number of geographically distributed databases. The algorithms presented here range from low granularity to high granularity decompositions of the algorithms.