The work deals with main aspects of computer efficiency increase. Object of investigation is a system consisting of processor, memory, other components and connecting links called buses. Rather different systems are used in modern world of information, so an interest arises to compare architectures of several producers. Comparison of systems is quite possible as the same architectural features bind them together: processor‘s clock speed, cache, memory, dual channel technology and others. To perform a comparative analysis, software has been used enabling to reveal increase of efficiency of separate computer components. Systems chosen for study are rather new from the point of view of technology: Intel Pentium 4, AMD Athlon XP and AMD Sempron. Experiments having been fulfilled, it came out that efficiency of a system for the most part depends on processor capacity: increase of its clock speed results in 9 – 13%, L1 cache has an effect even up to 1350% (theoretically), L2 cache – 30 – 38%. HyperThreading has been observed to mostly result in operations with floating point numbers (even up to 68%), and branch prediction would have theoretically to increase efficiency up to 47%. Estimating indicator of efficiency of the whole system, the results show that the main role belongs to processor. Influence of other components of the system is less noticeable. Working peculiarities of memory type determine rate of data selection and transmission from memory. The study has shown that... [to full text]

Duomenų bazių našumo analizė yra viena iš pagrindinių siekių programinės įrangos testavimo srityje šiuo metu. Atliekant tyrimus sukurta nemažai metodikų, kurios leidžia nustatyti duomenų bazių našumo lygį. Tačiau priemonės sukurtos remiantis šiomis metodikomis yra prieinamos dažniausiai tik uždaroms bendruomenėms. Šiame darbe yra nagrinėjama duomenų bazių našumo tyrimo metodika, pagrįsta programinės įrangos apkrovos ir stresinio testavimo principais. Sukurtas įrankis suteikia vartotojui galimybę atlikti duomenų bazių našumo tyrimo scenarijus, bei pakartotinai atlikti istorinius scenarijus ir palyginti gautus rezultatus. Formuojamos ataskaitos pateikia daug svarbios informacijos skirtos analizei. / The analysis of the database performance is the common challenge in the nowadays software testing. There are several methodologies of the analysis of the database performance in the market. However, tools, which are based on these methodologies, are available for the narrow circle of the privileged persons. According to the results of the analysis, this master thesis investigates a new methodology, which is based on several other methodologies. The methodology of the database performance audit, which is discussed in this project, is based on the principles of the software load and stress testing. In order to identify the issues of database performance, values of the performance parameters are registered. These values are counted during the execution of the scenario of the automatic scenarios. The user has the possibility to re-execute historical scenarios and to compare the results of the separate executions. Generated reports with the deep data level facilitate the analysis of the database performance.