Cut-through switching has attracted much attention as a promising approach to achieve ultra low latency in datacenter networks. Several cut-through switches that can forward a packet before its reception completes are already commercially available. However, despite the growing interest, the performance of cut-through switching in various network environments is not well understood. In this work, we propose a queueing-based framework to understand the behaviors of cut-through switching and to analyze its delay performance. Unlike traditional queueing models, we take into account the first-bit arrival of a packet as well as the last-bit departure of a packet to capture the essential features of cut-through transmissions.