Route-over VS Mesh-under

based on IP IoT architecture, technology and Applications

Interconnecting Smart Objects with Ip:thenext Internet

Adam dunkels (Adam Dankel) Jean-philippe Vasseur

The book mentions the problems of Route-over and Mesh-under, although it uses only three pages and does not involve too much technical detail, but explains the two routes from the most fundamental source. The contents of the previous see, although the details, but it is easy to confused people, do not know the relationship between the two.

Two-tier routing vs three-tier routing

The discussion about routing in the LLN network should be run on the second layer (link layer) or the third layer (network layer) has become a very sensitive issue.

Strictly speaking, routing means that the protocol and mechanism are to calculate the path on the third layer (IP) of the multi-hop network. It is also possible to perform a path calculation on the second tier of a multi-hop network using a MAC address, which is often referred to as "Mesh-under". In contrast to this is "route-over" (routing, so on the third layer).

Most people think that IEEE 802.15.4 will be the only low-power link available (always a risky assumption). When using a single link layer, the path calculation can be performed either at the link layer (the second layer) or at the IP layer.

But the advent of new low-power second-tier technologies strongly requires the use of layered architectures. This guarantees the independence of the layer, and in particular the second level of "agnosticism". Remember that the ability to use multiple link layers is one of the fundamental components of the TCP/IP architecture design. The routing protocol is obviously becoming necessary, which leads to the establishment of roll work and the design of RPL. (In fact, IP is intended to shield the underlying heterogeneous links and mechanisms)

The new question then arises as to whether a multi-tier routing architecture is required. Some link layer computed paths behave as third-tier IP links that perform routing operations across IP links.

The multilayer routing architecture is described in Figure 5.5. At the IP layer, the node performs the IP routing function, and the link layer node is not visible. At the link layer, the node uses the MAC address for "non-link" "Routing", which computes the path at the link layer. In this case, only N1, N2, and N4 are at the link layer.

First, such a multilayer approach is thought to be designed and deployed. The 6LoWPAN workgroup even defines a mesh address header for its 6LoWPAN adaptation layer, which supports the "Mesh-under" method by using the 802.15.4 address for each hop encoding, because it works on the link layer. No such link-layer routing protocols have been designed at this time.

Layered approach, the consequences of this routing architecture are well worth considering, and the book cited the consequences of two aspects:

One: the lack of visibility. Because the third layer considers the path of the second layer to be an IP link, the IP routing protocol is not visible to the link layer path. Inevitably leads to suboptimal routing. In fact, link-layer routing is based on its own weights and constraints to calculate the path, and the resulting path attributes are not notified to the IP layer. Such links have static weights that are independent or inconsistent with the IP layer. See the textbook for details

Second, network rerouting problem. The main attribute of a routing protocol is to look for an alternate path in the network during a network component failure (link or node). This involves both two-and three-tier simultaneous rerouting issues. See Textbook

Conclusion: With the advent of many low-power link layer technologies, such as IEEE802.15.4, Wi-Fi and PLC, network layer routing is obviously necessary. Although link-layer routing may be feasible on some link layers, the attempt to adopt a multi-tier routing architecture in a LLN network is obviously not a viable option given the dramatically increased complexity and inefficiency of the network.

For a mesh topology, here are a few things to know:

Can be consulted: RFC 6606 question statement and requirement For6lowpan routing

1, mesh is two layer, IP is three layer. Because the link layer address, or MAC address, changes after each forwarding (the physical address of the previous hop and the physical address of the next hop), the real destination address and source address are placed in the mesh header for transmission.

2, Mesh-under and Route-over is the difference, the former is two layers, is actually lower than the IP layer, the latter is the IP layer.

3, mesh of course is for the multi-hop, then in the Wpan network is not without IP, only with mesh can complete transmission? (this question should have been 3.1 solved, if the use of Mesh , only the border routers are IPv6 routers)

4, when using the mesh header, will bring some benefits. Because the interface identifier is stored in the mesh header, it is possible to compress the information in the IPV6 header, and the routing table does not have to store a 128 IPv6 address, but the 64-bit interface identifies the size of the routing table.

Here, "Routing" are not Equivalentto IP Routing, but includes the functionalities of path computation andforwarding t He IP layer.

The term "Routing" was used in thefigure in order to illustrate which layer handles path computation and packetforwarding I n Mesh-under as compared to Route-over.

Figure 1:mesh-under Routing (left) Androute-over Routing (right)

Route-over VS Mesh-under