Quartulli, Antonio and Lo Cigno, Renato (2011) Client announcement and Fast roaming in a Layer-2 mesh network. UNSPECIFIED.

Preview

PDF Download (898Kb) | Preview

Abstract

Research in Wireless Mesh Networking has been really active in the last years, with the birth of various ideas and the protocols that implement them. Nevertheless all the currently most deployed protocols have a lot of room for improvement and there are still open challenges that are meant to bring the performance to higher levels. Wireless Mesh Networks are involved in several studies also because of their applications which can be particularly useful in all those situations where wired infrastructures are not feasible. Common implementations rely on Layer-3 addressing and routing, but doing Mesh networking on Layer-2 address space is probably the best way for building a protocol needed to handle this kind of networks, since this task is not strictly related to the network (IP) layer. In particular, efficient client managing is not possible in case of a mesh routing protocol implemented on Layer-3 and this is probably one of the most important tasks of the routing protocol (after establishing routes, of course). This technical report focuses on the client announcement and the roaming problems that a protocol, embedded between Layer-2 and Layer-3, has to face. In particular, it focuses on B.A.T.M.A.N.-Advanced and defines new mechanisms that are meant to improve the protocol performance in client management and roaming. The current client announcement mechanism suffers from several problems, in particular from a high protocol overhead. The new strategy described in this technical report is meant to reduce such overhead and to increase the robustness of the whole announcement strategy. Up to now B.A.T.M.A.N.-Advanced hasn't ever provided a real mechanism to handle the roaming procedure, so this is the first one accounting this issue in the protocol. Performance are increased dramatically thus making it possible to measure real benefits in the user communications. This is probably the most prominent result obtained in this work which actually brings new possibilities for further improvements and for designing new mechanisms. The B.A.T.M.A.N.-Advanced protocol is actually transparent to the network layer by design, so the base idea behind the new roaming mechanism is to keep such transparency providing an on-time re-routing of the batman encapsulated data packets to the new destination. In this way it is possible to avoid affecting upper layers keeping them unaware of roaming, thus keeping all the connections active. The new mechanisms have been tested in the worst case roaming scenario for B.A.T.M.A.N.-Advanced. Such scenario consists in a topology made up by a chain of nodes of length $N$ (the longer the chain, worse the performance) on top of which a client served by the first node and one served by the last one are communicating. In case of any client movement (e.g. roaming to an adjacent node) the time needed by the new routing information to be spread among the network, and consequently make the connection work again, is proportional to the number of links that separate the two involved nodes. The experimental results, that are reported at the end of the technical report, show that even in this particular scenario the performance improvement given by the new mechanism is significant. The tests have been performed in an emulated environment where virtual machines, which acted as nodes, were created using QEMU-KVM. Therefore it has been possible to use the real B.A.T.M.A.N.-Advanced code instead of a possible error-prone implementation for any of the known network simulators (NS-2, NS-3, OMNET++, etc.).

Actions (login required)