Cascading failure is one of the key issues affecting the reliability of edge-assisted Internet of Things (IoTs), but is rarely studied. In this paper, a cascading model is developed with full consideration of the realistic characteristics of edge-assisted IoTs (i. e., congestion sensitivity and data convergence). In this model, the load of edge-computing nodes is represented by the real-time number of data packets, and is affected by the congestion state of links and the data-compressing ability of nodes. The experimental results have shown that being isolated is the main cause of performance degradation in edge-assisted IoTs during cascading failure; cascading reliability is positively correlated with link density and cluster head ratio; increasing the overload-tolerance coefficient and variable compression ratio can improve the cascading reliability of the network; decreasing the congestion-tolerance coefficient can reduce the damage of the cascading failure to the network; cascading failure triggered by high-directional betweenness attacks is the most damaging.