Information Technology Reference
In-Depth Information
So, how does this help? Well, suppose that, in the PhoneMob system,
changes in regulations governing the sending of phones to certain countries
result in a need to extend the list of repairOrder status values to include
awaiting export clearance. If we update some service centres before others,
the later ones may receive status messages with an unknown value. If, on the
other hand, we perform a two-phase update so that each service centre is first
updated to accept and store the new value but not to generate it, there will
be no problem with communication between new and old versions. Once this
change has been made and become stable, we can make a second change to
each system in turn so that they begin transmitting the newly defined value,
and the new functionality becomes effective.
Although this example is trivial, much more complex transitions can be
organized as a complex sequence of groups of updates, organized so as to
ensure that all subsystems are ready to handle new functions before they are
activated. This will generally involve quite a number of steps in order to
ensure that functional dependencies are observed. Checking that the update
schedule is sound and robust is a complex task, and is another area where tools
can be used to create and validate suitable sequences from the specifications
of the system before and after the change.
The computational language contains specific rules for supporting this kind
of evolution because it is by the creation of computational bindings that we
link subsystems that are at potentially different stages in the required sequence
of updates. However, similar analysis and reasoning can be applied in the other
viewpoints, although in the more abstract viewpoints we take a less localized
view of the system's behaviour and so just have models of the states before and
after the change. It is in the computational viewpoint that we introduce the
interfaces at which distribution may potentially take place, and so it is here
that the management of system evolution has to take account of interaction
between domains in which different sets of updates have been applied. It is
therefore natural that this is where the corresponding rules are positioned.
13.5 Evolution of the Enterprise
So far, we have considered incremental modifications. Some of the changes
to an enterprise and its infrastructure can be much more drastic. Consider
the large-scale changes involved in the big events in an organization's history,
such as mergers and acquisitions, or significant divestitures. At these points
there is a major upheaval in the provision of IT support, and sometimes this
severely damages the organization.
We can treat mergers in many ways as being similar to the establishment
of a federation. The key is to establish a clear enterprise model of the organi-
zations that are to merge, and use this as a reference throughout the process.
