• View on the problem
• New approach to integrated databank creation principles
• Conclusions

Nowadays almost all large projects in the field of information technologies stipulate creation of integrated databanks consisting of several databases with shared access tools. Relational database management systems (DBMS) predominate on the appropriate market. While being superb means to store, search, and retrieve static homogeneous data, these DBMSs are currently used in new application fields without proper analysis of conformity of data organization relational principles with nature of the data themselves. Those fields are first of all high-technology ones. They demand continuous use of most innovative techniques to cope with challenging problems and to ensure a competitive edge. It results in involvement of new data to be processed and analyzed.

One should note that in theory the relational model is able to describe any objects and phenomena of real world quite finely. But in practice, realization of dynamically interrelated data processing in real conditions and on real hardware engenders a lot of problems. The matter is in very principle of relational model. It is known to reject explicit preset relations of data objects. Those relations are established at the moment of data retrieval. It leads to excessive distraction of computer resources. Therefore, data model developers have to pre-arrange relational tables structure very carefully to minimize non-productive expenses of computational power and to attain required data processing efficiency. They must take into account all the range of tasks to be solved, since it determines involved data combinations and frequency of data queries. Thus, conventional scheme of integrated databank development, in general, consists of three consecutive stages:

• specification drafting of all the data to be stored;
• relational tables structures elaboration on the basis of the project tasks analysis and generalization;
• hardware and software implementation of created data model.

Unfortunately, economic and functional effectiveness of the project can be determined only upon accomplishment of third phase. Virtually inevitable miscounts at first two stages may seriously aggravate databank's functionalities. As a result, significant long-term capital expenditures may not be recouped. Customer's attempt to remove drawbacks results in never-ending iterative process of improvements. It demands new expenses and does not lead, as a rule, to desired result. During this time (usually several years) a customer is certain to be interested in new tasks. To solve them suggests new raw data structure which may greatly differ from original one. In such a situation customer is forced to choose one of the following:

• development of new databank project, which implies new expenditures;
• refusal to solve new tasks by means of created databank tools;
  
• customization of created databank to meet new demands, which leads to new relational tables generation and, as a consequence, to inevitable data multiplication.

Of course, none of enumerated decisions can satisfy a customer. Though third alternative seems to be acceptable at first glance, its repeated employment gives rise to rapid data multiplication. After having reached some critical point the scattered data consistency support becomes troublesome problem. Information used for decision-making loses its authenticity, and, as a result, validity of those decisions turns out to be questionable.

The authors offer full-scale realization of object-oriented approach to integrated databanks development. It is most prospective for dynamically grown spheres of human activities.

Quint-essence of this approach is concentration on the very data rather than tasks to be solved on their basis, because it is data that reflect outward things. Storage element in such a database (DB) is a set of object's parameters. It represents an adequate model of the object as applied to scope of involved tasks. Thus, key distinction from relational approach is a possibility of virtually unlimited build-up of object's features in DB. Moreover, and it should be stressed, object-oriented approach provides qualitative transformation of these features to relations with other objects.

Original object's description in an object-oriented DB by no means assumes any account of tasks the object will be engaged in. For example, an object may be simply empty and represented by its name only. This principle ensures most important properties of object-oriented approach, namely, dynamism and scalability. Represented by such a DB model of some human activity field keeps pace with the latter reflecting its innovations. Stored in DB objects acquire new parameters and relations as new data processing and analysis procedures appear, every new object's state inherently absorbing all previous ones. Data description tools in object-oriented DB permit objects to inherit and develop all acquired peculiarities. By the time of allocation in DB every object, like a germ, contains all possible properties and relations.

That is fundamental difference between object-oriented principle of DB construction and network one. The latter necessitates for all objects' relations to be predefined. Object-oriented approach provides more adequate model of outward things since it reflects their development, meanwhile relational and network approaches reflect only their state. This approach also separates data both from applications and tasks solved within those applications as well.

As a matter of fact, object-oriented approach unites advantages and eliminates drawbacks of both conventional and network approaches. From network principle it adopts direct establishing of objects' relations but does not require their static fixation. And like relational approach, it depicts an object as independent parameters set, but this independence does not impede imposition of any sophisticated relations between objects.

Authors have virtually tested considered approach as integrated databank development means for oil and gas industry enterprises, with its efficiency having been repeatedly proved. Above-mentioned principles of object-oriented databank creation and development enable one to realize a lot of databank functions on personal computer platform. Such a databank holds diverse information about thousands of oil-wells and includes specialized software means of data processing and analysis. At the same time, any application has direct access to relevant data allocated in DB, therefore, data multiplication is eliminated. Data retrieval procedures provide quick information query to solve quite complex tasks, for example, gridding, contouring, graphic representation of long-history statistic data, etc.

Practical use of proposed approach has given rise to innovative technique of databanks development. According to it, real databank creation may start from scratch. At first some small subsystem is accomplished to deal with limited tasks set. Nevertheless, the subsystem is capable to produce definite useful effect. Such a kernel may be used to model future databank functions. Obtained results are useful for readjustment of further development. As a result customer continuously has at his disposal efficient tools to resolve his tasks. He can put forward any considerations to improve the system functionalities. In addition, there is no need for large and risky capital expenditures, and every subsequent project development stage may be financed with use of returns that have been already brought by previous one. From economic viewpoint overall result is a considerable reduction of total expenditures and investment risks while from functional one - efficient tools for topical application tasks solving and well-founded decision-making.

1. Use of relational DBMSs for information technology projects development may be efficient mainly in those application fields which employ statically defined data sets and rather simple procedures of their processing and analysis, for example, bank operations, human resources accounting, trading, maintenance of archives and libraries, etc.
2. Fast-moving high-technology fields (oil and gas industry, remote sensing of Earth, environment monitoring, etc.) require permanent renewal and build-up of techniques for compute-intensive tasks solving to keep pace with progress. These tasks imply involvement of new and new data from adjacent spheres. Therefore, object-oriented approach to DB organization is reasonable here both from economic and functional viewpoints.