Smart grid communication networks adopt a variety of communication technologies interconnecting numerous and diverse equipment. The requirement of supporting a large traffic volume over such networks efficiently, reliably, and fairly among various applications calls for the study of the effectiveness of the transmission control protocol (TCP). The question raised in this paper is whether various TCP variants perform differently. We are particularly interested in comparing the performance of TCP-Reno and TCP-Vegas variants when handling sensory and metering traffic. The former variant follows a loss-based congestion control mechanism while the latter adds delay-based enhancements. Contrary to expectations, our simulation and analytical comparison results show that TCP-Reno outperforms TCP-Vegas in terms of packet loss rate. However, to take advantage of Vegas’ added features, a previously proposed scheme named split- and aggregated-TCP (SA-TCP) is recommended. The scheme splits meters’ TCP connections at intermediate devices and forwards collected data over an aggregated TCP connection to a data center.