Network dimensioning has been in the center of interest ever since the earliest networking architectures were put to use in the seventies. Since then service providers try to serve an ever increasing traffic with a given QoS. As a result, one of the main challenges of network dimensioning is how to design minimum capacity and minimum HW complexity network access modules which can carry a given amount of traffic with some predefined QoS parameters. The QoS is usually measured by cell loss probabilities and average cell delays. In this paper we provide dimensioning algorithms for Hierarchical Access Modules (HAMs) built up by subracks in a hierarchical manner and developed to provide different types of user-access to internet. The main objective is to design minimum complexity HAMs, which can serve a given population of users with a pre-negotiated level of QoS. Since the complexity and cost are measured by the number of racks and the link capacities used in HAM, optimal design is defined here as constructing a HAM which contains the smallest possible number of racks (cards) and the lowest capacity links. However, there are plentiful various topologies and corresponding link capacity arrangements which can serve a given population of users with a predefined QoS. As a result - when HAM is described as a multidimensional vector representing the topology the number of subracks and the corresponding link capacities - the optimal solution is to be sought in a large vector space. This casts dimensioning as a combinatorial optimization problem and we refer to this task as Node and Capacity Arrangement Problem (NCAP). In this paper, we develop such algorithms and provided numerical results for typical traffic loads and QoS parameters which can occur while using HAMs.