ROLAP

General information

Analytical requests in a relational database

Analytical requests in ROLAP are under construction over virtual multidimensional view of data, and their accomplishment happens at the level of a relational database, i.e. SQL queries over a relational system are executed. The main components of architecture of databases are the fact table (fact table) and dimension tables (dimension tables). The fact table is the main table of the database. It usually contains data on objects or events which set will be subjected to the analysis. Dimension tables contain constants or seldom impure data. They contain at least one descriptive field and an integer key field for unambiguous identification of the measured value. The dimension table shall be in the "one-to-many" relation with the fact table; if each measurement is in one dimension table, then such scheme is called "star"(star. If at least one of measurements is in several interconnected tables, then such scheme of creation is called "snowflake" (snowflake schema).

Application conditions

If the multidimensional model is implemented in the form of a relational database, it is necessary to represent it as long and "narrow" fact tables and rather small and "wide" dimension tables. Fact tables contain numerical values of cells of a hyper cube, and other tables define the multidimensional set of measurements supporting them. The part of information can be obtained using dynamic aggregation of the data distributed on the normalized structures differing in the architecture from "star", but in this case the requests including aggregation at the high-normalized structure of a DB can be executed quite slowly. Submission of the multidimensional information using star-shaped relational models fixes the problem of optimization of storage of disperse matrixes which is particularly acute for multidimensional DBMS in which the problem of sparseness is solved by the special choice of the scheme. Though for storage of each cell the whole record including except directly values is used, auxiliary keys  are links to dimension tables, nonexistent values just do not join in the fact table.

Quality evaluation

ROLAP system have the advantages and shortcomings in comparison with the multidimensional systems.

Advantages

• relational DBMS can work with very big DB and have the developed administration functions. When using ROLAP the size of storage is not so important parameter, as in a case with MOLAP
• at online analytical processing of contents of the data warehouse the ROLAP tools allow to make the analysis directly over storage, usually corporate data warehouses are implemented using relational DBMS
• at the changing dimension of a task when changes are made to structure of measurements rather often, ROLAP of a system with dynamic view of dimension appear as the best solution as in them such manipulations do not require physical reorganization of a DB.
• The ROLAP systems can function at much less powerful client stations as the main computing loading is the share of the server where the difficult analytical SQL queries created by a system are executed
• relational DBMS provide much higher level of data protection and good opportunities of differentiation of access rights

Shortcomings

• Limited opportunities of calculation of values of functional type.
• Smaller performance, than at MOLAP. For ensuring performance, comparable with MOLAP, the relational systems require careful study of the scheme DB and special setup of indexes. But as a result of such work performance of well customized relational systems when using the star scheme is comparable with a performance of systems on the basis of multidimensional DB.

See Also

• MOLAP
• OLAP

Literature

• Labotsky V.  V.  Knowledge management: technologies, methods and means of representation, extraction and measurement of knowledge  — Minsk: BGEU, 2006
• Alperovich M. Technologies of storage and processing of corporate data (Data Warehousing, OLAP, Data Mining)
• Shchavelev L.  V.  Methods of analytical data processing for decision support.  — Open systems,  No. 4-5, 1998

Links

• Internet university of high technologies
• SIT-forum
• Knowledge Management