Data centers are scalable and host a heterogeneous kind of applications and store an enormous amount of data. Various types of Data Center Network Architectures (DCNAs) are available for building a variety of Data Center Networks (DCNs). Multi-path DCNs allow servers to transmit and receive data through multiple paths. This helps to achieve better robustness, effective fairness, and high aggregate throughput. In this paper, a comparative analysis of several commonly used DCNAs is performed using scalability, path diversity, latency, throughput, cost, and power consumption metrics. Single-path Transmission Control Protocol (TCP) is the backbone of the Internet. TCP offers reliable data delivery and effective congestion control capabilities. Variants of TCP ensure that TCP works efficiently in a WAN. But due to the single-path transport, TCP is incapable of efficiently utilizing high-bandwidth and low-latency DCNs. This paper also analyzes incompatibility issues between the single-path TCP protocol and multi-path DCN. Further, we evaluate the performance of single-path TCP by running it over a multi-path FatTree DCN. Simulation experiments are performed using the ns-3 simulator to validate our analysis. Goodput and delay results are collected by configuring an increasing number of pods within the FatTree DCN.