This paper is a research of problem which was faced during development of clinical information system. This is a scope, where one clinical test can by characterised by several hundreds of attributes. Thus the focus is on the design and performance of database tables with 200-300 columns. To perform the research the Oracle Database Management System was chosen. There is the main factors, influencing performance, such as principles of data storage on disk, query processing also the main facilities to improve performance observed. Based on the basic principles of relational database design some alternatives to tables with many attributes were created. It is normalized and denormalized structures. There is performed the experiment to see how the speed of most used queries depends on data structure. Also there is done the analysis of complexity and usability of these queries.

Šiame darbe atliekami stochastinių minimumų asimptotiniai tyrimai. Įrodomos minimumų ribinės teoremos tuo atveju, kai tiesinis normalizavimas neduoda neišsigimusių ribinių skirstinių, tokiu atveju taikau netiesinį minimumų normalizavimą. Konkretaus skirstinio atveju randamos netiesinės normalizavimo funkcijos, kurių pagalba yra gaunami minimumų klasikiniai ribiniai skirstiniai. Įrodoma Perkėlimo teorema netiesiniam normalizavimui. Darbo tikslai: • ištirti netiesinio normalizavimo reikalingumą; • išanalizuoti netiesinio normalizavimo galimybes minimumų schemoje. Darbo uždaviniai: • parinkti netiesinio normalizavimo funkciją konkretaus skirstinio atveju; • gauti ribinius klasikinius skirstinius, kai minimumai normalizuojami netiesiškai; • įvertinti konvergavimo greitį ribinėse teoremose; • atlikti aproksimavimo paklaidų kompiuterinę analizę. / This paper is the asymptotic analysis of stochastic minima. Proofs of minima limit theorems are provided for cases, when linear normalization does not give non-degenerate limit distributions. In this cases, non-linear minima normalization is used. For a specific distribution, non-linear normalization functions are calculated, which are then used to get classic limit distributions for minima. Objectives: • Examine the necessity of non-linear normalization; • Analyze the possibilities for non-linear normalization in minimum pattern. Tasks: • Choose non-linear normalization function for a specific distribution; • Get classic limit distributions, where minima are normalized non-linearly; • Investigate the rate of convergence within the limit theorems; • Perform computer-based analysis of approximation errors.