Database Reference
In-Depth Information
the trajectory reconstruction process explained in Chapter 2 is part of the ETL
process in a TDW architecture.
Continuing with Figure 4.1 , a DW makes use of metadata, which include
information about the DW schema, the data source schemas, the mappings
between source and DW attributes, as well as the frequency of data refreshment.
From the organizational DW smaller DWs can be built to satisfy departmental
needs. These DWs are called data marts .
On the next tier, an OLAP server provides a multidimensional view of the
data stored in the DW. This enables analysts, managers, and executives to gain
insight into data through interactive access to a wide variety of possible views
of information. Thus, at a conceptual level , data are perceived by the user as
a hypercube where each cell contains values, called measures , which quantify
facts . The axes of the hypercubes are called dimensions . Dimensions are typically
organized into hierarchies , which allow to aggregate measures at different levels
of detail. Queries addressed to the OLAP server are expressed using OLAP
operators such as slice, dice, roll-up, and drill-down. The slice operator removes
a dimension in a cube, that is, obtains a cube of n 1 dimensions from a
cube of n dimensions. This is analogous to a relational algebra projection. Dice
applies a Boolean condition to a cube, and returns another cube containing only
the cells that satisfy such condition. This is analogous to a relational algebra
selection. Roll-up aggregates measures according to a dimension hierarchy,
using an aggregate function, to obtain measures at a coarser granularity. Drill-
down disaggregates previously summarized measures, and can be considered
theinverseofaroll-up.
Finally, the user interacts with the OLAP server through several tools, such
as OLAP, reporting, statistical, and data-mining tools. In the case of an OLAP
client, the user can then perform OLAP analysis interactively.
If a DW stores trajectory data, we are in the presence of a TDW. Typical
analysis over a TDW includes finding out the distribution of trajectories by road
type (which requires a roll-up operation to aggregate trajectories by road type,
and a slice operation to keep the dimensions of interest), or the total number of
cars in a certain location at a given moment. We will give examples of TDW
queries in the following sections.
At the logical level , a typical implementation, referred to as relational OLAP
(ROLAP), stores the data in relational databases. This leads to two kinds of
tables. Fact tables store the data elements under analysis (e.g., trajectories in
a TDW), while dimension tables describe the axes of analysis (e.g., roads,
vehicle type) of the data contained in the fact tables. If dimension tables are
denormalized, that is, there is a single table for the whole dimension, we
have a star schema . Otherwise, that is, if there is one table for each level in
a dimension hierarchy, we have a snowflake schema . Fact tables are usually
normalized.
Search WWH ::




Custom Search