Add project "Mesh channel coordination"

Signed-off-by: Vincent Wiemann <[email protected]>

collections/_projects/mesh-channel-coordination.md

@@ -0,0 +1,125 @@
+---
+name: "Mesh channel coordination"
+desc: "Automatically select channels in mesh networks"
+requirements:
+ - "C programming skills"
+ - "Experienced Linux knowledge (usage and architecture)"
+ - "WLAN technology"
+ - "Game theory"
+ - "Graph theory"
+difficulty: "high"
+
+mentors:
+ - "Vincent Wiemann (CodeFetch)"
+initiatives:
+ - GSoC
+ - ffh
+ - freifunk
+tags:
+ - OpenWrt
+ - WLAN
+ - Networking
+ - Mesh technology
+ - Game theory
+ - Graph theory
+ - Linux
+ - C
+ - Lua
+collaborating_projects:
+ - "freifunk"
+ - "ffh Freifunk Hannover"
+---
+
+The aim of this project is to improve the quality of the wireless user
+experience in mesh networks and to increase the stability of wireless mesh links
+in general specifically mitigating the hidden node problem by utilizing
+diverse channels where adjacent networks might otherwise interfere.
+
+At the moment most mesh solutions define a default channel value for their
+wireless mesh networks. In many cases the nodes have additional means to connect
+to their neighboring nodes like direct wired connections or VPN connections.
+Some nodes utilize multiple radios to reduce the hidden node problem with
+diversity routing.
+
+Upcoming features like MCCA, BSS coloring and dedicated RUs will further allow
+reducing contention caused by interference.
+These also need a form of coordination which has been subject of current
+research and which is partly being defined in the WiFi 7 standard.
+
+As a mesh network is a highly dynamic system and as the radio is often used to
+both serve the clients and forward mesh traffic in parallel (so-called VIFs -
+virtual interfaces), the configuration of the channels must be done respecting
+the special properties of mesh protocols while taking route switches into
+consideration and reacting accordingly. Care must be taken to not influence the
+routing decisions of the mesh protocol in a way that leads to an imbalance.
+
+An example for an imbalance occurs in dense high-use scenarios with packet-
+loss-metric based mesh routing protocols if a wireless link appears to be
+saturated due to interference caused by neighboring nodes and their clients.
+If a redundant route exists in that case, nodes might choose another route with
+a presumably better metric.
+In theory that switch was assumed to be transparent to the user, but in practice
+a mediocre link is often swapped for a link that e.g. has even less capacity
+congesting the link long enough for some clients to assume a broken connection
+and switch to another node which will eventually suffer, too.
+This problem can be mitigated using adjacent channels where coverage overlaps
+and a redundant connection exists.
+
+Even though modern mesh routing protocols are hardened against some of these
+issues by not only considering the packet loss for their metric calculation but
+also other criteria like calculating the route metrics from wireless statistics
+exported from the kernels mac80211 subsystem, they are still an open topic of
+research.
+
+For example the highly sophisticated diversity routing RFC amendment of the more
+and more popular routing protocol Babel does not consider shared usage of the
+radios' mesh network with the client network which can lead to unexpected
+behavior as different WLAN clients tend to prefer quite individual frequencies
+depending on their hardware, antenna and software configuration which could
+favor an imbalance.
+
+As this project involves a lot of different disciplines the main goal of the
+student is not to get deep knowledge about every internal. Instead, the student
+shall be encouraged to think and act independently then cooperatively and to
+find a simple, creative way to improve the current situation.
+
+This reflects the results-oriented approach to work which is often required for
+participating in open-source projects where code changes must be kept simple to
+be reviewed timely and to increase the chance of being accepted.
+
+
+#### Ideas
+* Create "hearing maps"
+* Find ways to create hierarchical sub-trees in the network by identifying
+  "border nodes" (e.g. VPN or PtP links) which is critical for distributing
+  the coordination
+* Only use the links which are actually needed: If a node can reach all of
+  its  neighbors by other means (e.g. VPN) it should not be bound to a specific
+  channel assuming the available bandwidth offered by the alternative link and
+  the metric is sufficient. It may even be possible to deactivate the wireless
+  mesh network for that radio to save airtime (e.g. if there is no neighbor).
+  This needs to dynamically react if e.g. a new neighbor appears.
+* Analyze the roaming behavior of the clients to estimate where interference
+  due to the hidden node problem is more likely.
+* Take actual traffic flow patterns into account.
+* Identify e.g. long-range interference using coordinates if given similar
+  to the WiFi 6 proposed Automatic Frequency Coordination scheme.
+* Make use of game theory to achieve a Nash equilibrium instead of central
+  coordination.
+* Where a form of centralization is unavoidable, make the consensus fault-proof.
+  The system must reconstitute timely if the elected coordinator fails.
+  The Raft consensus protocol is an interesting candidate for this.
+* Allow manual override
+
+#### Milestones
+* Add basic support for mesh networks in DAWN
+* Implement a daemon "meshrealmd" which can interact with DAWN and implements
+  the graph theory-based approach presented in the paper "Interference-Aware
+  Channel Assignment in Wireless Mesh Networks" in a decentralized way
+* freifunk-gluon integration (OpenWrt-based mesh framework)
+
+#### Inspirationals
+ * https://blog.freifunk.net/2018/05/19/dawn-a-decentralized-wifi-controller/
+ * https://link.springer.com/chapter/10.1007/978-3-540-89183-3_7
+ * https://dblp.org/pid/175/1661.html
+ * https://github.com/willemt/ticketd