 |
OLAP: The Panacea for the Ills of
Management Information Systems ?
|
 |
Other OLAP Resources
Links: Definitions of OLAP, Features of an OLAP database,
OLAP Products.
The term OLAP (On-Line Analytical Processing) was coined by
Dr. E.F. Codd, his wife S.B. Codd and their associate, C.T.
Salley in the research paper "Providing OLAP to
User-Analysts: An IT Mandate". This paper was published in
early 1993 and was sponsored by Arbor Software Corporation, the
creators and vendors of ESSBASE.
In the early 1980s, Dr Codd laid down rules which formed the
basis on which the current generation of relational database
technology is built. In the early 1990s, Arbor Software were
looking to identify a niche market which would justify the
co-existence of multi-dimensional and relational databases. The
research work which Dr Codd and his colleagues carried out has
resulted in a formalised re-definition of the requirements for
tools to implement decision support, management information,
business intelligence and executive information systems. Tools to
support business modelling will be obliged to pay attention to Dr
Codd's rules in order to remain at the forefront. The publication
of the paper has already caused significant changes in the market
place. Most notably to the benefit of Arbor Software who have now
formed strategic alliances with several existing players in the
marketplace and whose products were rumoured to be on line to be
distributed by Microsoft in a deal which will make a
multi-dimensional engine available behind Excel.
In general terms, the controversy over the rules has mirrored
that which surrounded the publication of the relational rules.
Product vendors have either been keen to prove how compliant
their tools are and to extend the rules to reflect their own
tool's strengths or have poured cold water on them, arguing that
the research was biased and unnecessary.
As with many dichotomies, truth probably lies somewhere
between the two extremes.
What is an OLAP database?
Links: Top of Document,Features
of an OLAP database, OLAP Products.
"...a multidimensional server database which makes
available management information interactively to a client on the
end-user desktop. The client can be a spreadsheet-like graphical
interface, a bespoke interface, an executive information system
or any other similar tool. In all cases, users can choose the
information they want using simple English language dialogue
boxes and navigation tools, changing the data, with immediate
response as they work. Users can carry out their own analysis in
accordance with their individual needs and IT departments can
stop worrying about user requirements."Computing,
9 June 1994
This is obviously a wide definition which covers a range of
applications and tools. The requirements of the above definition
could also be meet by a range of different types of data storage
and presentation technology from flat files to relational
databases. It is also very much a technical definition.
Dr Codd defines OLAP as "... the name given to the
dynamic enterprise analysis required to create, manipulate,
animate and synthesise information from ... "Enterprise Data
Models"... This includes the ability to discern new or
unanticipated relationships between variables, the ability to
identify the parameters necessary to handle large amounts of
data, to create an unlimited number of dimensions (consolidation
paths) and to specify cross-dimensional conditions and
expressions." "Providing OLAP (On-Line
Analytical Processing) to User-Analysts: An IT Mandate",
E.F. Codd, S.B. Codd and C.T. Salley, 1993, courtesy of Comshare.
He goes on to outline three significant characteristics
of OLAP:
- Dynamic Data Analysis,
- in which historical data must be manipulated extensively
involving multiple data dimensions. It can provide an
understanding of the changes that take place within an
enterprise over time. Several product vendors have
concentrated on producing databases to support this
"time series" analysis of data.
- Four Enterprise Data Models.
- These four models allow for the definition of data
structures (Categorical), the storing of historical data
in the structure (Exegetical), the exploration of
"what-if" scenarios (Contemplative) and a model
to indicate the complex relationships which exist between
apparently disconnected variables (Formulaic).
- Common Enterprise Data.
- The data which is presented through any OLAP access route
should be identical to that used in operational systems.
Features of an OLAP Database
Links: Top of Document,Definitions
of OLAP, OLAP Products.The
following points summarise the key aspects of the 12 OLAP rules
as defined by Dr. Codd in his paper.
- Multi-Dimensional Conceptual View
OLAP databases support a multi-dimensional view of the
data allowing for the classic "slice and dice"
operations or pivoting and rotation of the conceptual
cube of data. This might involve looking at the data in
terms of the products or product categories on display as
well as by the outlet channel that provided the business
and then moving on to levels of persistency inherent in
the business from each source. All of this available as
and when required by the user-analyst. Interestingly,
this rule is given subtle levels of shading by some
suppliers of OLAP-type software who argue that a
multi-dimensional conceptual view of data can be achieved
without multi-dimensional storage.
- Transparency
The users should have no need to know that they are
looking at an OLAP database. As far as they are
concerned, they are using tools with which they are
familiar to get the data they require in order to make
the decisions they have been charged with making. Nor
should they need to know the source of the data. For
example, there should only be one definition of
persistency and this should be applied at all data
sources irrespective of their provenance.
- Accessibility
The tools in use should have a map of data sources within
it (the implementation of the Categorical model) which
point it to the most appropriate source of data to
support a specific query and perform any conversions of
data or semantic meaning in order to give an agreed and
predetermined interpretation of the enterprise business
model.
- Consistent Reporting Performance
As the number of dimensions or the number of levels of
aggregation changes, there should be no alteration in the
way key figures are calculated. The system models should
be robust enough to cope with changes to the enterprise
model. This is essential if the figures presented in the
OLAP tool are to be believed and its analysis or
predictions are to be trusted.
- Client-Server Architecture
The OLAP tools should be capable of being deployed in a
client-server environment, implying that the
multi-dimensional database server should be accessible
from a range of other applications and tools. This is
probably one of the most contentious of Codd's rules
since few products in the marketplace currently meet this
criteria. ESSBASE and, to a lesser extent, Express are
probably the only two that can comply.
- Generic Dimensionality
"Every data dimension must be equivalent in both its
structure and operational capabilities ... The basic data
structure, formulae and reporting formats should not be
biased toward any one data dimension." "Providing
OLAP (On-Line Analytical Processing) to User-Analysts: An
IT Mandate", E.F. Codd, S.B. Codd and C.T. Salley,
1993, courtesy of Comshare.
- Dynamic Sparse Matrix Handling
Typical multi-dimensional models can easily run into
millions of cell references, many of which will have no
appropriate data at any one point in time. These null
values should be stored in an efficient way and not have
an adverse affection the accuracy or speed of information
retrieval.
- Multi-User Support
OLAP tools should support and indeed encourage group
working and the interchange of ideas and analyses between
users. To achieve this, multi-user access to the data is
essential.
- Unrestricted Cross-Dimensional Operations
The rules which govern the progress of data "roll
ups" through levels of a hierarchy should be defined
and available so that no matter which slice of data is
taken, the rules will be applied consistently.
- Intuitive Data Manipulation
The user-analyst's view of the data should at all times
contain all information necessary to effect the
navigations (the slicing and dicing) which are
appropriate without the need to resort to the use of a
menu or multiple trips across the user interface.
- Flexible Reporting
The user should be able to retrieve any view of the data
required and present it in any way that they require.
- Unlimited Dimensions and Aggregation Levels
There should be no limit imposed by the OLAP tool to the
number of dimensions which can be built into a model.
Links: Definitions of OLAP, Features of an OLAP database, OLAP Products.
Goto Top, Goto Main Page.
e-Mail: simong@sgroves.demon.co.uk