Abstract
Fast data distribution is important for many businesses and services today. If data distributions like operating-system deployment, media/file distribution and patching is slow it could negatively impact productivity. The work in this thesis is aimed at improving performance of data distribution where the receiving nodes are in a data subscriber relationship. The main idea is that this could be achievable by first discovering the network topology and then traverse the network with a snake like behavior, in other words traversing each network link only once. In this thesis the focus is on exploring what performance gain there could be when using the proposed distribution strategy.
The thesis goal is approached through two main steps. First, transport protocols are used to benchmark switch duplexing performance. These benchmarks are aimed at finding out what performance can be expected and if the proposed distribution strategy is viable. Second, a proof of concept prototype based on the proposed distribution strategy is created. The prototype is compared with BitTorrent in a distribution scenario. The findings in this thesis show that there could indeed be a performance gain by using the proposed distribution strategy.