Prioritization of HPC Projects in the Public Sector: A Decision Support Framework Based on the FTOPSIS-Class Method

This paper presents a framework for the prioritization of projects that require high-performance computing (HPC) infrastructure in the public sector. In view of the scarcity of video card (GPU) resources in the sector, in addition to the complexity of evaluating the level of prioritization that each project should receive, this project addressed the use of the FTOPSIS-Class multicriteria decision method to order requests from public agencies to have access to GPUs based on criteria such as social impact, cost reduction, efficiency expectation and level of technical maturity of the applicant. The method uses fuzzy logic to address the uncertainty inherent in the qualitative judgments of  plaintiffs, who, in turn, may not have the quantitative certainty of what they are requesting. Experiments with five different sensitivity scenarios, two of which are anchors, demonstrated the robustness of the model by adapting the priority queue according to government strategic guidelines, keeping it aligned with the expected prioritization levels. The results obtained validate the system as an effective tool to improve technological governance and transparency in the distribution of limited state resources.