Fairness is a topic that emerges in many fields and that is linked to resource allocation and fair division problems. In networking and computing the legacy approach to solve these situations is to model them as a single-decision maker problem, using classical resource allocation protocols as the proportional rule or the max-min fair allocation. The evolution of telecommunication network technologies, the advances in computing power and in software design practices allow giving a high degree of freedom and programmability to resource allocation and routing decision-making logics. Furthermore, software-defined radio and virtualized network platforms can be used on top of a shared infrastructure making possible a real-time auditability of the system by its tenants and users. Therefore, novel networking contexts such that tenants can be aware of other users’ demands and the available amount of the resource, or they can have a partial information on the system, have to be considered. Moreover, in the decision-making modeling for 5G systems, it is necessary to move from single-resource allocation to multi-resource allocation. In fact, with the introduction of network slicing, we need logically-isolated network partitions that combine network, computation and storage programmable resources. In this thesis we aim to provide a theoretical and formal analysis and redefinition of fairness of resource allocation for congested networked systems, i.e., systems that are in the challenging situation in which resources are limited and not enough to satisfy users’ demand. We analyze, propose and evaluate numerically centralized, decentralized, single and multi-resource allocation rules.