Search

In distributed databases, oriented toward general business software such as groupware, the asynchronous update method is widely used in order to align the databases. In this method, each database has replicas of the other databases, and the replicas are updated at the same time according to a predetermined replica routing schedule. This paper discusses an approximate method for minimizing the communication cost and time needed for updating the replicas in such a distributed database. More precisely, the case in which the communication cost is proportional to the amount of transferred data is considered, and an optimization problem is formulated in which the replication routing schedule minimizing the time for the replica update is to be determined from those minimizing the communication cost for the replica update. An approximate solution for the formulated problem is presented. Then the result of evaluation of the approximate replication routing schedule obtained as the approximate solution is discussed, and it is shown that the communication cost for replica update is minimized and that the time for the replica update is at most 1.2 times the minimum value if the number of distributed databases is at most 7. © 2001 Scripta Technica, Electron Comm Jpn Pt 1, 85(4): 61–70, 2002 [ABSTRACT FROM AUTHOR]

Distributed databases in general-purpose business software such as groupware widely use a method for simultaneously updating the replicas of other databases owned by each database according to the replication routing schedule. In this paper, a replication routing schedule is discussed in which the time required for updating the replica is the minimum in the distributed database. Specifically, an optimization problem is first formulated for deriving the replication routing schedule with minimum update time. Next, in the cases of even and odd numbers of distributed databases, an algorithm providing a replication routing schedule as an optimum solution is presented in the even case, while a heuristic algorithm providing an approximate solution is shown for the odd case. These algorithms are extremely simple and are practical since the required amount of computational processing is small. © 1998 Scripta Technica, Electron Comm Jpn Pt 1, 82(2): 11–20, 1999 [ABSTRACT FROM AUTHOR]