Physical Data Warehouse Design - Data Warehouse Systems: Design and Implementation - page 274

when all possible views are materialized. Also, we showed that when full
materialization is not possible, we can estimate which is the best set to be
chosen for materialization given a set of constraints. We then studied two
typical indexing schemes used in data warehousing, namely, bitmap and join
indexes, and how they are used in query evaluation. Finally, we discussed
partitioning techniques and strategies, aimed at enhancing data warehouse
performance and management. The last three sections of the chapter were
devoted to study physical design and query performance in Analysis Services
and Mondrian, showing how the theoretical concepts studied in the first part
of the chapter are applied over real-world tools.
7.12 Bibliographic Notes
A general topic about physical database design is [ 116 ], while physical design
for SQL Server is covered, for instance, in [ 35 ]. Most of the topics studied
in this chapter have been presented in classic data warehousing papers.
Incremental view maintenance has been studied in [ 73 , 74 ]. The summary
table algorithm is due to Mumick et al. [ 141 ]. The PipeSort algorithm, as
well as other data cube computation techniques, is discussed in detail in [ 2 ].
The view selection algorithm was proposed in a classic paper by Harinarayan
et al. [ 78 ]. Bitmap indexes were first introduced in [ 149 ] and bitmap join
indexes in [ 150 ]. A study of the joint usage of indexing, partitioning, and view
materialization in data warehouses is reported in [ 12 ]. A topic on indexing
structures for data warehouses is [ 98 ]. A study on index selection for data
warehouses can be found in [ 60 ], while [ 192 ] surveys bitmap indexes for
data warehouses. A popular bitmap compression technique, based on run-
length encoding, is WAH (Word Align Hybrid) [ 232 ]. The PLWAH (Position
List Word Align Hybrid) bitmap compression technique [ 192 ] was proposed
as a variation of the WAH scheme, and it is reported to be more ecient
than the former, particularly in terms of storage. The authors proposed this
indexing scheme to support music data warehouses, which we comment in
Chap. 15 . Rizzi and Saltarelli [ 177 ] compare view materialization against
indexing for data warehouse design. A survey of view selection methods
is [ 128 ]. Finally, [ 10 ] discusses practical aspects of the Mondrian OLAP
engine.
7.13 Review Questions
7.1 What is the objective of physical data warehouse design? Specify
different techniques that are used to achieve such objective.
7.2 Discuss advantages and disadvantages of using materialized views.