up the appropriate subschema. Name of the field,
that stores the value, is not the name of the attribute;
3. Data structure in the database is represented
by a sparse matrix. Logical data structure is formed
actually, i.e. comprises only entities and attributes,
that are really used in system, and can be
supplemented or changed at any time;
4. Total number of tables in a database on the
physical layer is fixed and depends only on the
number of metadata tables and the types of data and
is independent of the logical structure of data stored.
Limitation of the number of tables in the physical
layer is the key to increase the productivity and
reduce the complexity of database queries and
maintaining the characteristics obtained during its
operation.
These requirements are fulfilled by most of
known models of structure-independent databases
that use relational technology (Paley, 2002; Tenzer,
2001; Polishchuk, Chernykh, 2009), but each model
corresponds to a specific motivation of the
developer.
3 THE METHOD OF
CONSTRUCTING THE
STRUCTURE-INDEPENDENT
DATABASES
Regardless of the motivation of the developer there
is a finite sequence of steps leading to obtain the
necessary set of linked relational tables
corresponding to above requirements. This set of
tables is the physical layer structure-independent
database, and the sequence of steps is a method of
constructing a structure-independent database using
relational technology (Rogozov Y., Sviridov A.,
Kucherov S., 2014).
For the first requirement it is necessary:
1. Determine the set of metadata needed to
describe the logical structure of the database;
2. Identify the metadata structure and implement
it in the form of a set of linked relational tables that
make up the metadata subschema. This step
includes:
•divide of generated metadata set into groups of
tenancy;
•create for each such group a separate relational
table;
•specify the relationships between tables;
To fulfill the second requirement it is necessary:
3. Determine the types of data that will be used
to store the attribute values;
4. Develop data subschema in the form of a
finite set of similar unrelated relational tables needed
to store attribute values. It is advisable to store the
values in the form of triples "entity- attribute-value",
where the first two elements – are links to entries in
the metadata tables. This enables to present the data
structure in form of sparse matrix;
For the final formation of structure-independent
database, you should:
5. Implement the physical structure of the
database.
Performing the third and fourth requirement is a
consequence of implementation of the first two
requirements:
• Presentation of stored data structures in the
form of sparse matrices is implemented by storing
the values in the form of triples (sparse matrix of
instances) and storage of metadata within the
database (sparse matrices of entities and attributes);
• Independence between the physical and logical
levels is implemented by storing metadata in the
database.
Application of the above sequence of steps
precedes the development of user (logical) data
structures (Rogozov Y., Sviridov A., Kucherov S.,
2014), which is performed by the developer-defined
database technology.
4 THE PROOF OF METHOD
Proposed method allows to get different structure-
independent databases using relational technology
depending on the goals of the developer. To prove
this statement, let’s consider the example of
obtaining the already known structure on the basis of
this method.
Example 1. In (Paley,2002) authors describe the
database model, which is the aim of developers to
create a universal relational structure for processing
and storage of quasi-structured data in terms of
object-oriented technology – in the form of classes
and objects. According to the proposed method of
structure-independent database designing to obtain a
model of such structure is as follows:
Stage One. Defining Metadata Sunschema.
Step 1. To store entities and attributes, it is
needed to select the representation format
"Directory". In this case, the implementation of
the entity set
i
eE
is represented by the
following ternary relation (table):
R
E
),,( FEIDE
ER
DATA2014-3rdInternationalConferenceonDataManagementTechnologiesandApplications
140