Temporal View Self-Maintenance in a Warehousing Environment

Abstract

Warehouse view self-maintenance refers to maintaining materialized views at a data warehouse without accessing source data. Self-maintenance has been studied for nontemporal views, but is even more important when a warehouse stores temporal views over the history of source data, since the source history needed to perform view maintenance may no longer exist. This paper tackles the self-maintenance problem for temporal views. W e show how to derive auxiliary data to be stored at the warehouse so that the warehouse views and auxiliary data can be maintained without accessing the sources, and we give algorithms for tal self-maintenance of all warehouse data. The temporal view self-maintenance is considerably harder than the nontemporal case because a temporal view may need to be maintained not only when source data is modified but also as time adv ances, and these two dimensions of change interact in subtle ways. W e must take both dimensions and their potential interactions into account. W e also seek to minimize the amount of auxiliary data required, taking into account different source capabilities and update constraints that are common in temporal warehousing scenarios. While our framework and algorithms are presented using a temporal data model, our results apply directly to the adhoc temporal support (i.e., timestamp attributes in the standard relational model) commonly found in data warehouses today .