The ODP reference model work has identified 5 viewpoints which may be
used to examine a distributed system. These viewpoints are tools to
help with the analysis of a distributed system. Different groups of
people from different cultures are involved in specifying, building,
using and running a distributed system.
These viewpoints represent the important perspectives of the distributed
system to different groups of people involved in the system.
Each group sees some aspects in great detail, but other parts are
described in the vaguest way as they are not important to that group.
For instance, the users of the system will want to know a lot about
what the system does in terms of using information and producing
useful results, they will not be interested in exactly which CPU
model, or disc drive, or even disc sector, they are using to achieve
their work.
On the other hand the operations manager will be very concerned about
CPU utilization, disc blocking, and sharing hardware resources, he
will not be very interested in what the distributed system is actually
doing in terms of the input and the results.
The five viewpoints are not meant to represent 5 layers, however they do have
some relationship, in that we can see how refining one viewpoint may
well lead to another. Viewpoints are not in and of themselves refinements or part of any particular development methodology.
All view points are always valid.
The viewpoints will overlap, there is no distinct boundary.
The viewpoints are just perspectives on an underlying model, or actual
system. If there were no underlying model then the perspectives become
disjoint. When used in the context of an underlying model then each perspective
(viewpoint description) is different in the concepts that it brings into
focus as important and the concepts that are hidden or ignored.
Each perspective will represent a different level of abstraction of
the system. These are illustrated in table#tbvp#49>.
A number of projects have contributed to the
the way these ideas may be projected onto distributed systems in five
ways:
-
Enterprise
This projection describes how computing systems and the distribution
of computing resources match the objectives and organization of the
enterprise. This shows what the system's function is within the
enterprise, rather than how it functions.
Note that ;SPM_quot;enterprise;SPM_quot; represents any human endeavor in a social
context. It does not just mean a company, it might be a trade
association, a charity, a standards group, community of interest, anything that requires
interaction and computerization.
In this viewpoint, distribution is inherent, but largely transparent.
-
Information
This projection identifies the information used within the
organization. It shows where the information is used, and in which
ways it is processed. It indicates the meaning of the information within the
system to its users.
An information model reflects the real world, for instance a stock
record should be matched by actual stock on shelves, as
accounting information is matched by money in the bank. Workflow
analysis is often used to derive the information model implicit in an
organization.
The information projection does not just contain the information
requirements and models of the enterprise, but also the information
requirements of the distributed system. For instance for security
access control information must be kept and this can be described
from this viewpoint. Obviously the access control information will select the
enterprise model for access control, but is part of the information
needed within the distributed system.
Thus, there are two components of the information viewpoint:
information requirements for the enterprise, and for the distribution
itself.
Distribution is usually taken for granted in this viewpoint, so it is
inherent and transparent.
-
Computational
This projection shows how the applications may be structured so that
they are independent of the computer systems and networks on which
they run. It produces a computational model for :- how information and
its processing are represented, how applications are specified and how separate
applications are linked and interfaced. This viewpoint would include any
inherent distribution and parallelism in the application.
The importance of this viewpoint is that it represents the
capabilities of the distributed system to the applications programmer.
In particular, it enables the programmer to invoke all the
transparencies that are required from a distributed system (see
below).
Alternatively, of course, the programmer can ignore the transparencies
and handle distribution directly.
So in this viewpoint distribution can be transparent, or it can be
made explicit.
-
Engineering
This projection relates the underlying components of a distributed system
to the applications. It is here that such issues as performance,
reliability and availability are decided.
This is the viewpoint where distribution is explicit, and must be
handled. For instance, replication mechanisms might be introduced to
increase reliability. Different mechanisms would be selected based on
engineering tradeoffs concerning cost, performance and availability
constraints.
-
Technology
The technology projection maps the distributed system onto the
processing nodes and communications infrastructure that form the
hardware and software base.
Distribution is not part of this viewpoint.
#table52#
Table: Framework of Abstractions document
Who uses which viewpoint is show in figure {ref<#58#>tb:vpu<#58#>.
#table59#
Table: Viewpoint users
<#229#>#tex2html_wrap3794#<#229#>
The Open Distributed Processing model is still undergoing evolution.
What is presented in this book is not meant to reflect that
communities model at all, but this author's extraction of those points
that were found useful, and reasonably complete at the time of
writing.
<#230#>#tex2html_wrap3796#<#230#>