On the Effects of Maximum Transmission Unit in Power Line Communication Networks

Abstract

Currently, in many power line communication networks, the maximum packet size for the PLC protocol used to send data, in the interest of obtaining the maximum possible throughput. However, electrical noise on the power lines can cause poor performance for large packet sizes. In this paper we investigate this problem by building a network measurement tool, called GatorBytes to measure throughput and packet error rate over power line communication (PLC) networks. We found the packet error rate as to be as high as 78% on a low throughput link and around 44.8% on high throughput links. We suspected that the high packet error rate was due to large packets on the wire because large packets tend to have high probability of error resulting in packet drops due to noise interference. Large maximum transmission unit (MTU) may result in high throughput on a clean channel; however, it may cause low throughput on a noisy channel. To remedy this problem in PLC networks, we advocate the idea of lowering the MTU size dynamically to optimize throughput. To support this proposition, we conducted several throughput and packet drop rate measurements over PLC networks by using commercially available PLC devices and variable message sizes. We examined issues such as the MTU size of PLC devices, the relationship between packet error rates and packet intervals, and the relationship of packet size and throughputs that have not been previously analyzed. Our experiments showed that we were able to obtain 39% performance gain on aggregate throughputs and up 78% improvement in packet drop rate, by using lower MTUs.