<p>Network tomography (NT) aims to estimate the states of network links from end-to-end path measurements. Two major categories of methods have been proposed to address the issue of link states inference, known as Analog tomography (AT) and Boolean tomography (BT), respectively. In this paper, we developed a fairly general approach called sequential network tomography by exploiting the temporal correlation of link state. For the widely used AT and BT methods, we have proposed their respective sequential versions. In particular, as the temporal correlation of the link metrics between adjacent time slots exists, the estimated link states in the current period provide a solid foundation for making more accurate predictions in the subsequent period. We perform simulation experiments for delay estimation and congestion identification across a range of network topologies, demonstrating that the proposed sequential-based scheme can achieve better performance than conventional sparse based network tomography.</p>