Krzysztof Kaczmarski
PhD thesis, June
Methodologies and Metamodels
for Object-oriented Grid Databases
Summary
In this thesis we describe a way to
design and create object-oriented grid databases. The task of transparent
integration of distributed resources has waited many years. Nevertheless there
are many prototypes, experimental scientific and commercial systems, it may be
noticed that existing technologies do not offer any solution, which could
satisfy all potential groups of users. Such a system must work in such a way, that is maximally transparent not only for a client but
also for a programmer and an administrator. Recently, grid technologies made
significant improvements in the field of running distributed jobs. The way of
their operating (improving comfort of work in distributed environments),
which they utilise
may be applied to databases resulting in unusual effects. A grid database
offers a completely new way of using a distributed computer system. Users think
they do not work with many systems, differing in any possible aspect, but they
see all of them as a single, complete virtual repository. It is called virtual because it
does not exist in any particular place but is simulated in a moment, when it is
needed. It does not use resources more than those that are necessary to perform
such a simulation and it does not store state what improves and simplifies its
modification.
Creation of such a virtual
repository seemed to be impossible in the field of object-oriented databases. However,
evolution of programming languages, network technologies and mostly invention
of Stack-Based Approach with updatable views, has changed this situation
dramatically. However, operation of such a system based on probably many
related views raises new challenges. How such a system may be modelled? How can we express the ways of data integration?
Is it possible to design and deploy such a system using existing CASE tools?
This thesis gives answers to these and many other questions. It describes not
only a prototype system, which transparently integrates resources, but also a
way of modeling and creating such an integration. The
concrete results of this work are:
• modification of some SBA concepts to fit into
distributed environment;
• working prototype of a grid database
which integrates distributed resources using updatable views;
• general sketch of multi layered grid
architecture with methodology of its creation;
• analysis of requirements for designing
integration realized by updatable views;
• metamodel created upon above analysis and
suited to model grid database;
• UML profile which allows modeling of
a grid database using existing CASE tools.
Right now, the author under eGov-Bus project realized among others by Polish-Japanese School
of Computer Science (within 6th EU Framework Programme)
participates in development of a new database system, which will combine
properties of the described prototype
and a professional object database with
complete programming capabilities. There is a chance that the problem of
transparent integration of distributed databases will be finally solved.