Problem of effective data processing in scientific applications is well-known. There are two reasons of that: nonoptimal front-end application code and ineffective system of scientific data management at back-end. This article presents differences between flat file and relational database data storage. It also shows (based on Microsoft SQL Server database engine) benefits and methods of taking advantages of relational database. These include creation scientific workgroup environment for high availability, fault tolerance and optimal scientific data processing with build-in implemented structures and mechanisms. Such solution can be useful in numerical methods programs, also in Finite Element Methods.