From d62329ab09b60d61b114ef1e7dbd60ff4aaefb79 Mon Sep 17 00:00:00 2001 From: Richard Whitehouse Date: Thu, 19 May 2011 14:43:20 +0100 Subject: [PATCH] Conclusion draft --- conclusion.tex | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) diff --git a/conclusion.tex b/conclusion.tex index 4d8a1ef..44ab529 100644 --- a/conclusion.tex +++ b/conclusion.tex @@ -1,2 +1,27 @@ \chapter{Conclusion} +\section{Future Work} + +Based on the work outlined in this dissertation, there are a number of areas of possible enhancement. + +\section{Dynamic Routing} + +At the moment this dissertation relies on static routing, based on a each switch being told of it's neighbours. While in some situations this is a possibly, by directly managing the switch, it would be useful to have the switches dynamically configure themselves. + +By incorporating the SAMI protocol outlined in \cite{dwh}, and the Rapid Spanning Tree Protocol \cite{ieee802-1d}, it would be advisable to create a Bridge Protocol Data Unit, BPDU, based protocol to provide for this. Once specified, this could be incorporated into the simulation to provide additional metrics for comparison. + +\section{Spanning Tree for MOOSE} + +By using the link state data further, it would be possible for the MOOSE protocol to be smarter about how it forwards broadcast packets. Currently these are handled by using Reverse Path Forwarding, having, as shown in the evaluation this causes additional traffic, that does not occur under RSTP, as in RSTP, each broadcast packet is sent along more links than are necessary. However, the MOOSE switches could potentially use a spanning tree of the switches to provide for a minimum spanning tree implementation. + +\section{Summary} + +In this dissertation I have succesfully implemented a working implementation of MOOSE, and Ethernet for a modern network simulator, ns3. This has involved teaching myself a large number of technologies, MOOSE, the workings behind ns3, the waf build system, the detailed implementation of Ethernet, to support knowledge I have learned elsewhere in the Tripos, such as in Principles of Communication. I have expanded on my knowledge of using C++ to create a system of significant size which can be used by other people for research in the future. + +The MOOSE implementation is fully functional, remembering state information, rewriting packets, including ARP requests and responses. I have designed experiments to show that my switch works under a variety of conditions and have included information to show the differences between the protocols. These included data on the different properties of the routing protocols, with MOOSE outperforming Ethernet in terms of unicast routing, while Ethernet performs better than MOOSE under broadcast. + +I have also shown how the different protocols cause different sizes of forwarding tables in the routers, with MOOSE performing drastically better than Ethernet under some situations, while Ethernet performs marginally better than MOOSE in others. + +I have also made suggestions for how the MOOSE protocol may aim to overcome some of the limitations identified, by extending the routing protocol to improve broadcast performance. + +The insight I have gained into the MOOSE protocol and ns3 has allowed me to suggest further developments for the protocol, beyond the scope of this dissertation. I also aim to have the code produced in this dissertation pushed upstream into ns3 to allow other developers to expand on the work outlined here. -- 2.34.1