Page 38 - Building Digital Libraries

Choosing a Repository Architecture

Lower cost. Rather than buying hardware to accommodate the
maximum anticipated use, you can provision only what is
necessary and provision more or less as needs dictate.
Avoid vendor hardware lock-in.

On the same hardware, virtual machines are inherently slower because the
emulation layer consumes resources. In addition, services hosted by virtual
machines in shared environments may affect each others’ performance.
However, most repositories will perform well in a properly managed virtual
environment.

What Data Model Does Your Repository Need?
A data model is an abstraction that organizes elements and standardizes
how they are related to each other. For purposes of creating, maintaining,
and migrating repositories, it is important to be aware of the following:
Data models are only relevant to the extent that you can organize
resources and their associated metadata into a form that
supports the functionality you need, and you can export this
same data in a manner that can be made understandable to
future systems.
Data models do not define the contents of elements, how systems
interact with elements, or any other aspect of system behavior.
As a result, the underlying data model is only one factor of
many that determine whether a system meets a need.
Most systems have their own data models, and as of this writing, no
widely supported data model yet exists. Even when different
systems have the same data model, there is no guarantee that
they can share information.
Data and resources can usually be migrated from one data model
to another, particularly when the migration is across systems
that are designed to perform similar tasks.
A system may be fully compliant with a data model even if it
doesn’t claim to be so.
A system may meet both user and library needs even if it does not
support the desired data model.
For the purpose of understanding data models, consider the Portland
Common Data Model (PCDM). The PCDM is a simple but flexible model
specifically created for repositories in that it is designed to express structural
relationships between objects, as well as provide access control. A few things
are particularly interesting about PCDM:

• It formalizes structures that can be found in systems that
have been around for many years—including paper-based
ones that predate computerization.

23