Zbog ubrzanog razvoja informacijske tehnologije su nastali složeni sustavi kao što su računarstvo u oblaku. Navedeni sustavi najčešće moraju imati visoku razinu dostupnosti podataka, odnosno moraju osigurati neprekidni rad poslovnih sustava. Kako bi se to postiglo prisutni su visoki kapitalni i operativni troškovi podatkovnih centara koji su neophodni za ovakvu vrstu usluga. Mnogobrojna istraživanja na ovu temu ukazuju kako su poslužitelji glavni uzročnici visokih troškova podatkovnih centara. Upravo se zbog toga njihovi resursi nastoje što učinkovitije iskoristiti. U istraživanju su navedene Web-farme kao primjer iz prakse koji potvrđuje da postoje sustavi čiji poslužitelji nedovoljno iskorištavaju svoje resurse, ali ih moraju imati kako bi osigurali visoku razinu dostupnosti sustava. Spomenuti visoki troškovi i nedovoljna iskoristivost postojećih računalnih resursa su glavna motivacija za ovo istraživanje. Nakon proučavanja dosadašnjih znanstvenih istraživanja, ali i rješenja iz prakse, utvrđeno je da ne postoji rješenje koje bi dovoljno učinkovito riješilo ovaj problem. U radu se predlaže novi model za automatiziranu i poboljšanu iskoristivost postojećih računalnih resursa bez potrebe za ponovnim pokretanjem poslužitelja koji rješava navedeni problem. Na temelju modela je napravljena aplikacija koja je validirana na primjeru Web-poslužitelja gdje je ovaj problem prepoznat. U radu se koristi istraživačka paradigma znanost o dizajnu (engl. Design Science Research Methodology, DSRM), koja se temelji na kreiranju novog artefakta što u ovom slučaju predstavlja novi model.
Information technology is under constant innovation pressure to provide the highest level of data availability, i.e. the continuous functioning of operating systems. This is the very reason for an accelerated development of complex systems called cloud computing. One of the tasks of such solutions is to ensure high-level availability of complex systems and architecture. In order for such solutions to function properly, the high capital and operational costs of data centers are essential for this type of service. There are numerous studies which indicate that servers are the main cause of data centers’ high cost. As a result, the aim is to use servers, i.e. their resources more efficiently. This paper shows the examples from practice which confirm that there are systems whose servers insufficiently exploit their resources, but they must have them due to their importance. A concrete example of this problem are the Web Farms where, in order to achieve greater system availability, there is a greater amount of resources than is really needed, as confirmed by tools for measuring server loads. This approach allows the system to withstand sudden loads, which increases the level of system availability. The negative effect of such an approach is the increase in capital and operating costs due to a higher amount of computer resources. The mentioned high costs and inadequate utilization of the existing computer resources are at the same time the main motivation for this research. To solve this problem, it is necessary to have a system which would automatically allocate as much computer resources as the system needs, depending on its load and thereby taking into account its availability and consistency. During a detailed study of the current scientific research, as well as practical solutions, it has been found that there is no effective solution to this problem, which also served as an additional motivation for this research to be carried out. The existing solutions are lacking in that they are not dealing with how to use the existing resources more efficiently, but in adding new or migrating virtual servers to other physical servers in critical situations, which requires even larger numbers of computer resources. The second approach to solving this problem is process prioritization, i.e. that servers with the greatest need for resources are given the highest priority in the execution of the process. The disadvantage of this approach is that resources cannot be increased nor decreased, but only prioritized, which still results in the presence of unused resources. One of the disadvantages of the existing solutions is that it is not possible to add and subtract computer resources (CPU and memory) without the need to restart the server. A large number of existing solutions focus only on CPU or memory, but not on both. Due to all this, a decision was made to build a new model for an automated and improved utilization of the existing computing resources. The model will be verified by building an application that will also serve for validation on the Web server example where this problem was recognized.
The research paradigm used in this research is the Design Science Research Methodology (DSRM), which has specific guidelines for evaluation and iteration within research projects. The methodology is based on the creation of a new artifact. In this case that is a new model which addresses these complex problems mentioned in this case. The Design Science Research Methodology consists of six sequential process steps, which are: identification of problems and motivations, a definition of goals, design, and development, presentation of solutions, evaluation and communication. Throughout these steps, numerous methods and techniques were used such as: comparison, evaluation/validation, content analysis, experiment, modelling techniques (UML), diagram techniques (causal relationship diagrams), structural analysis of processes (decomposition diagrams, data flow charts, and block diagram), programming (pseudocode and scripting languages (BASH and PHP), as well as many others. With regard to scientific contributions, this research has resulted with a new model for an automated and improved utilization of the existing computing resources without the need to restart the server, as well as in clearly defined cases and constraints regarding the new model’s application. The research has shown that the application of a new model enables a more efficient utilization of the existing computing resources (CPU and memory) without the need to restart the server. The research also provides recommendations for the implementation of the model in the selected programming language, and the process of evaluating the model in the experiments. In view of the social contribution, the whole solution is open source, which is also one of the main goals of this research. This results in an easier application of the solution and the repeatability of the testing to facilitate further improvement and research on this topic.