Revisit on View Maintenance in Data Warehouses

Abstract

The complete consistence maintenance of SPJ-type materialized views in a distributed source environment has been studied extensively in the past several years due to its fundamental importance to data warehouses. Much effort has been taken based on an assumption that each source site contains only one relation and no multiple appearances of a relation is allowed in the definition of views. In this paper a generalized version of the view maintenance problem that not only a relation may appear many times in the definition of the view but also a site may contain multiple relations is considered. Due to unpredictability of the communication delay and bandwidth between the data warehouse and the sources, the materialized view maintenance is very expensive and time consuming. Therefore, one natural question for this generalized case is whether there is an algorithm which not only keeps the view complete consistent with the remote source data but also minimizes the number of accesses to the remote sites. In this paper we first show that a known SWEEP algorithm is one of the best algorithms for the case where multiple relations are included in a site. We then propose a complete consistency algorithm which accesses remote sources less than n - 1 times for the case where multiple appearances of a relation is allowed and n is the number of relations in the definition of the view.