They
can be categorized as refreshing with accessing data sources (e.g., Ceri & Widom,
1991) as well as self-maintenable refreshing (e.g., Samtani, Kumar, & Mohania,
1999). In the latter case, additional data structures are stored in a DW along with
materialized views in order to eliminate the need of accessing data sources.
The process of refreshing a materialized view is usually executed concurrently with
transactions on data sources and with user analytical queries. Such concurrent executions
may result in inconsistent data stored in materialized views and erroneous
results of analytical queries. Multiple solutions for avoiding these problems have
been developed, that is, recomputing a view from scratch, applying compensation
algorithms, maintaining versions of views, using additional data structures and
transactions (e.g., Gupta & Mumick, 1999; Quass & Widom, 1997; Zhuge, Garcia-
Molina & Wiener, 1996). Yet another problem is related to maintaining consistency
of multiple dependent views during the process of their refreshment (e.g., Colby,
Kawaguchi, Lieuwen, Mumick, & Ross, 1997; Zhuge, Wiener, & Garcia-Molina,
1997).
OLAP applications analyze data by means of complex queries ranging from a few
to dozens operations of joining, filtering, grouping, and aggregating. Since these
queries are very complex and they often read terabytes of data, their execution may
take dozens of minutes, hours, or even days.
Pages:
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26