From Conventional to Multiversion Data Warehouse - Evolving Application Domains of Data Warehousing and Mining - page 58

unIverSIty dAtA WArehouSe:
A cASe Study

Over time changes & problems: As a part
of making a common education standard in EU,
KTH has raised its Master's degree requirement
from 60 European credits (ECTS) to 90 ECTS. At
the same time, credits of each course are changed
from 5 to 7,5 ECTS. After a while, a semester was
divided into two learning periods.
The multidimensional schema of the university
data warehouse shown in figure 10 cannot adapt
new instances i.e. changes in degree requirements
from 60 to 90 ECTS cannot be stored in the data
warehouse. Also, degree requirement was raised
from 60 to 90 credits in 2007, a change valid for
all the students admitted from 2007 onwards.
However, if the degree requirement is raised in
the data warehouse, this will wrongly apply to
students from prior years. This is due to the reason
that old information is overwritten by new infor-
mation ( information loss problem ). Overwriting
of old information takes place because the data
warehouse schema has a static structure in what
concerns the schema, and it is not flexible enough
to store two instances of data. Similarly, in the
case of redesign of the university's semester plan,
when semester is divided into two learning period
(LP's), it requires changes in the dimensional
schema of the data warehouse. However, this
schema change cannot be handled by conventional
data warehouses.

In this section we present a versioning case study
for illustration purposes. The Royal Institute of
Technology (known as KTH) is a leading tech-
nology institute in Stockholm, Sweden. It has a
number of Schools and each School consists of
departments. Undergraduate, graduate and post
graduate programs are offered by each depart-
ment. School of Information and Communication
Technology (SICT) is situated in Kista Science
City and it has two major departments, Micro-
electronics and Information Technology (IMIT)
and Computer and Systems Science (DSV). A
program, Masters in Nanoelectronics is offered
by IMIT, whereas Masters in Information Sys-
tems Engineering (EMIS) is one of the 60 credits
degrees offered by DSV.
Figure 12 shows a dimensional schema of
the university data warehouse. The schema has
five dimensions, {Program, Student, Courses,
Time and Employee} and a fact table {Uni-
versityFact}. The dimension named 'Program'
has three hierarchal levels, at highest level the
university has 'Schools'. A school consists of
'Departments', which offer different programs.
Each department has employed teachers who
teach different courses, which are offered in
different semesters.

Table 2. Comparison of SQL extension and transparent querying approaches

Query extension approach Transparent querying approach

Query writing becomes very complex when versions are in-
creased.

Query writing is simplified even in the presence of large number
of versions.

By default partial results are produced. Partial results cannot be produced.

Results can be merged by using a clause 'merge'. By default, results are merged.

A set of versions can be included in the result. Data from all the active versions is produced and it is not possible

to limit the result to certain versions.

Can be used by expert users only. Can also be used by non-expert users.

Querying single/current version is possible. It is not possible to query single/current version.