Determination of buffer size for access networks in first generation optical networks

Abstract

A collection of computers linked together through a communication channel are known as a computer network. Such a network facilitates communication and resource sharing between netwo rks. The linkages are facilitated by data transmission lines, which can be optical or Ethernet cables . The data switching device can be a switch or a router. In times of delay or congestion in the network , routers hold packets in buffers . They do so before s ending the packets to the next destination. The size of a buffer has a significant impact on the performance and design of a router. The performance issues are related to network jitter and the physical size of routers due to buffer assignment standards. Over time, it has been proven that congestion in a network is in the access or edge network and not the core of the network. In addition to the traditional assignment of buffer size in a network, which is backed by Request for Comments and router manufa ctures, there have been alternative models proposed for the assignment of buffer size, but no buffer size is suitable to work for the whole length of the network. In this study, Riverbed Academic Modeller was used, to come up with a suggested model for t he calculation of buffer size for low and high traffic throughput. This model does not use link utilization as a metric for assignment of buffer size. In suggesting the model, the study establishes that for an access network of not more than 48 nodes, 32 p ackets and 28 packets are suitable buffers for a high and low throughput network respectively. Keywords: Internet, Transmission Control Protocol, Buffer Size, Datagram Forwarding Rate.