In this paper, we present a novel approach for resource provisioning in Time-Domain Wavelength Interleaved Network (TWIN), an optical burst switching solution available for transport in regional networks, with nodes equipped with transponders composed of a fast tunable transmitter and a fixedwavelength receiver. The approach consist in allocating the number of transponders to satisfy the traffic demand, providing at the same time an optimal slot schedule for the transport of optical bursts. Apart from calculating the optimal schedule, the solution also allocates the additional slots to sources, in order to improve the level of QoS latency experienced by individual network flows during the pre-insertion queueing process. This functionality is achieved by original QoS constraints. The dimensioning method is optimal, formulated as Mixed Integer Linear Program (MILP), and it allows a trade-off between the amount of network resources allocated and the QoS performances of the optimal solution. By simulations we studied the impact of the proposed dimensioning solution on the network performances and the total network capacity achievable by TWIN. We show that our proposed latency constraint mechanism leads to insertion latency limited to few time slots only at no additional network resource cost, in a 6node mesh network. We also show that non-uniform network traffic decreases TWIN capacity efficiency (i.e., wastes resources) compared with deterministic traffic.Index Terms-Optical burst switching, network dimensioning, QoS.