Repository : ssh://git@open-mesh.org/doc
On branches: backup-redmine/2017-07-13,master
commit 107b362e4ec446685ab421e7bc968f567170178a Author: Linus Lüssing linus.luessing@c0d3.blue Date: Tue Mar 9 00:20:08 2010 +0000
doc: open-mesh/Gsoc2010-ideas: adding ideas list for the GSoC 2010
107b362e4ec446685ab421e7bc968f567170178a open-mesh/Gsoc2010-ideas.textile | 64 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 64 insertions(+)
diff --git a/open-mesh/Gsoc2010-ideas.textile b/open-mesh/Gsoc2010-ideas.textile new file mode 100644 index 00000000..17df3b55 --- /dev/null +++ b/open-mesh/Gsoc2010-ideas.textile @@ -0,0 +1,64 @@ += Requirements for students = + * Excited interest in Mesh networking technologies + * Routing and networking knowledge in general + * Programming language: C += Recommended and/or useful = +These things are not a prerequisite but might be very useful and/or have to be learned turing the GSoC anyway. + * [[http://www.kernel.org/doc/Documentation/CodingStyle Linux kernel coding style]] + * Reading the documentation of the algorithms: [[http://gitorious.org/batman-adv-doc batman-adv-doc]] [[http://tools.ietf.org/html/draft-wunderlich-openmesh-manet-routing-00 batman-rfc-draft]] (used in the 'old' batmand) + * Being able to use virtual machines for debugging (i.e. qemu) + * Being able to install a mesh network with OpenWRT and batman-adv + * Kernel coding experiences (especially with how to handle and avoid memory management/race conditions/deadlocks) + * monitoring packet flow (using tcpdump / wireshark + [[http://gitorious.org/wireshark-batman-plugins/wireshark-batman-adv wireshark-batman-adv dissector]] + += Ideas = + +== link layer fragmentation / compression == +'''Brief description:''' Introducing link-layer fragmentation and header compression to offer alternative packet overhead solutions. + +Due to the packet encapsulation performed on each and every payload packet going through the mesh a small overhead is introduced. This overhead (24 bytes at the moment) must be dealt with by the admins deploying the batman-adv mesh. The most common solution is to increase the maximum packet size on all mesh interfaces to maintain the standard MTU of 1500 bytes on all client devices. While this works in most cases it does not work for everyone. Some devices / drivers don't support or allow bigger packets. Alternatively, each client has to reduce its MTU to 1476 bytes which is often not manageable. This project shall address the issue by either compressing the payload packets (e.g. VJ compression) and/or implement a leightweight link layer fragmentation and/or develop another solution. + +== forward error correction == +'''Brief description:''' Forward error correction to avoid retransmissions in the mesh network. + +Wireless networks rely on a lossy transport medium - even under optimal conditions packets get lost and have to be retransmitted. This comes at an even greater cost since the air is a shared medium (only one participant can transmit data at any given time) retransmissions have a severe impact on the network's performance. In addition to the packets which have to travel over several hops towards their destination, higher layer protocols (e.g. TCP) expect ACK packets to travel back over the chain of connected mesh nodes otherwise the transmission starts anew. This project aims to avoid retransmissions by implementing a packet forward correction for the batman-adv payload packets. A parity packet would be injected into the packet flow which allows the receiving batman-adv node to calculate & deliver the lost packet without further retransmissions. The parity packet rate can by dynamically adjusted depending on the link quality towards the final destination. + +== B.A.T.M.A.N. protocol overhead reduction == +'''Brief description:''' Split the currently used OGM packets into two seperate types to reduce the amount of packets flooded in the neighborhood. + +The traffic B.A.T.M.A.N. IV generates to discover the best path through the network is quite low compared to other protocols, but especially when B.A.T.M.A.N. has many single hop neighbors which rebroadcast each others OGMs we see room for improvements. The project shall optimize the flooding algorithm by splitting the originator message into two different message types. The OGM will remain but only be used to flood the TQ throughout the network. The new message type (a name needs to be found) will contain the link qualities of the single hop neighbors only. This message won't be rebroadcasted and just reaches the local neighborhood. These local message can be sent much more often than the global TQ messages and thus reduce the traffic [nearly just create a linear growth of traffic with more nodes in the local neighborhood instead of a squared amount]. + +== B.A.T.M.A.N. protocol convergence speed == +'''Brief description:''' Implement a fast changing environment detection algorithm to let mobile nodes adapt to their new surroundings faster. + +Fast moving nodes always have the problem of adjusting their routing information in time. They can choose to send more routing information, so that their environment can adjust to them but stationary nodes won't do the same and increase the mobile node's adaption time greatly. However, when a B.A.T.M.A.N. node detects that his local environment changed quickly he will enter the starvation mode. In this mode the node will actively try to confirm a working route as fast as possible by sending a "batman ping" to its new neighbors. Each B.A.T.M.A.N. neighbor will try to forward the message to its destination, once arrived there it will travel back. If the mobile node receives the reply it can change its route towards the new neighbor without waiting for normal OGM flooding as the route has been verified. The goal of this project is to implement the starving mode together with the "batman ping". + += Optimize multicast performance = +'''Brief description:''' Avoid broadcasting multicast traffic to nodes not belonging to a certain multicast group. + +Multicast packets would have a great advantage especially in wireless mesh networks: The number of recipients of multicast packets (i.e. for zeroconf service announcements or audio/video streams) is a lot higher then with unicast packets. batman-adv currently handles multicast packets the same way as broadcast packets, they simply get flooded through the mesh network. Instead, those packets should be flooded on a subgraph containing the nodes of a certain multicast group and other connecting nodes only. + += Dynamic OGM intervals = +'''Brief description:''' A batman-adv node shall select originator interval rates according to the dense and dynamics in its closer environment. + +Depending on the usage scenario, people can adjust the bandwidth being used for batman-adv's route finding algorithms. Usually people were advised to increase the originator interval if the mesh network is small but needs fast route refresh rates or to decrease it if the mesh network is mostly a static setup with a lot of nodes. It would be great if a batman-adv node could determine the dynamics of the mesh network it is currently participating in on its own so that this option would not have to be adminstrated from a person anymore. For instance in combination with the 'B.A.T.M.A.N. protocol overhead reduction' project, the local broadcasts could be automatically increased if there are not that many direct neighbours. + += More batctl ping/traceroute options = +'''Brief description:''' Add useful options for adminstrating a mesh network to batctl. + +batctl is the nicer fontend for configurating and debugging batman-adv. There are still some options that could be added like showing the MTU of hops with traceroute along a path to make it easier to spot MTU issues. Or outgoing interface selection to be able to manually probe the way and quality of an alternative path. Feature proposals that make the administration of a complex mesh network more easy are welcome. + += Live VIS in map = +'''Brief description:''' Building tools that visualise the dot output of the vis server with additinal gps coordinates on a map. + +With every new technology, a bridge to non-technical people should be provided as well. batman-adv is being used in routers of every-day users, that do not have an insight in the B.A.T.M.A.N. routing protocol itself, nevertheless a good visualisation can widely increase the acceptance of a new technology and get young people interested in it. batman-adv has a built in vis-server which produces a raw dot-file when activated. With the help of graphviz-tools, those dot-files can be rendered as graphs which are still more interesting for 'technical' people. It would be great to have a tool that maps the information provided by the dot-output and additional geo coordinates in Google-Maps or OpenStreetMap (OpenLayers) in realtime. Then 'normal' people could then find out and solve dead zones without technical support all on their own without having to be able to use fancy command line tools. This feature would be useful for anyone administrating (parts of) a mesh network. + += Multiple interfaces per node support in Mesh3D = +'''Brief description:''' Adapting Mesh3D to handle the new visualisation format features from current batman-adv. + +Mesh 3D is an applicaton written in C and maintained by one of the batman-developers which is able draw a 3 dimensional graph from batman's vis output in dot-format. The latest additions in batman-adv's vis ouput that now features a differentiated visualisation, in that interface connections between nodes are now being shown seperately. This new format feature has not been ported to Mesh3D yet. Also a concept for visualising overlapping links in Mesh3D would probably have to be planed (adding transparency to Mesh3D for instance). + += Multiple switch ports for redundancy = +'''Brief description:''' Allow multiple bridge uplinks to wired networks + +It is often desired to have multiple uplinks to a (wired) switched network where the B.A.T.M.A.N. protocol is not used. This may be a data center or a core network where multiple or redundant connections are needed. However if the Mesh network device and Ethernet device are bridged on multiple nodes, bridge loops are created. Traditional measures like (R)STP don't help in this situation as they may disable the (good) Ethernet Links. A solution should be developed where we can use multiple (redundant) uplinks to the same core switch network while effectively avoiding switch loops. \ No newline at end of file