Abstract:
Site multihoming refers to a situation when a site has access to the Internet through more than one Internet service provider. In the current Internet, multihoming is achieved by using Border Gateway Protocol. A multihomed site acquires its provider independent or provider aggregatable address block and then announces this address block through all its service providers to the Internet’s global routers. This solution works for IPv4, but scalability is an issue for IPv6. High demand for multihoming and the huge address space provided by IPv6 requires a solution which is able to scale well. A wide range of solutions have been proposed for IPv6 multihoming during past years. These solutions can be put into three categories: Routing approaches, Middle- Box approaches and Host-centric approaches. The first part of our research focuses on an analysis, from a deployability viewpoint, of seven proposed solutions which are still under investigation by researchers and Internet experts. The analysis shows that there is no perfect solution yet and all proposed solutions still need to be improved. However, host-centric solutions are more attractive from the viewpoint of deployability since they do not need any change in the Internet routing system. SHIM6 and MultiPath TCP (MPTCP) are two incrementally deployable solutions belonging to this category. Failure detection and recovery without breaking active communications is one of the main requirements for every multihoming solution. MPTCP and SHIM6 are both able to provide this functionality. SHIM6 is equipped with a specific failure detection and recovery protocol (REAchability Protocol) and MPTCP employs a central congestion control mechanism (SEMICOUPLED Congestion Control Algorithm) for this purpose. These protocols are not yet widely deployed on the Internet, so the efficiency of these protocols is still an open research area. The second part of the work presented in this thesis focuses on this research area by conducting a series of real-world and simulation experiments for SHIM6 and simulation experiments for MPTCP which have been designed to investigate the efficiency of REAP and SEMICOUPLED. Traffic Engineering is one of the major weak points of SHIM6. Since the solution is implemented in end-hosts, having a site-aware traffic engineering mechanism is challenging. To address this issue, we propose a dynamic traffic engineering mechanism which offers a rich set of traffic engineering features for SHIM6-enabled sites.