Inside Mesh Networks

Ad-hoc wireless mesh networks will be the great enabler for the IoE. Part one discusses the technology behind them.

Mesh networks have a huge upside when it comes to the Internet of Everything, but there are also some big issues that have to be resolved.

“One of the real challenges with mesh networks is there is not a lot of control of the devices that are joining and leaving the network,” says Steven Woo, vice president of enterprise solutions technology at Rambus. “A second challenge is that certain mesh networks are ‘spontaneous’ or ad-hoc, which will be very applicable to the IoE.”

But exactly how this ad hoc platform is going to work is still a bit murky. Ad hoc mesh networks popping up whenever and wherever can be an extremely disruptive element in carefully designed wireless networks, especially with the IoE—and especially if they become wide area. They also can play havoc with smaller areas such as a home or office, where other technologies such as Bluetooth, ZigBee, Ant and ZWave are present. A lot has to be resolved before ad-hoc mesh networks will play nicely with other existing wireless.

What can go wrong
The most prominent issues have to do with nodes that are either hidden, or which slip in and out of the net (edge issues). Figure 1 shows the footprint of three close proximity nodes, where only node B can see both A and C. This issue is the same whether the nodes are stationary or slip in and out of the network at the edge or when powered up. Node C has to contend with both nodes A and B, and can do so with various cooperative methodologies between the node that is online at the moment. But because nodes A and B are unaware of each other, they don’t consider each other when it comes to transmission protocols.

Each is only aware of A-B, or B-C communications. This can lead to various problems when node B is communicating with node A, and node C wants to link to node B, and vice-versa. The end result is that hidden nodes are often unable to hear each other. One way of looking at this is with the parallelism of a group of people communicating with each other in an ever ever-expanding circle. As the circle expands the people at the edge get farther and farther away, making it harder to hear them.

As a result, nodes can’t see or acknowledge each another, which means they won’t back off to allow others to “talk.” Hence only the strongest nodes, or the ones closest to the access point, get to communicate. Theoretically, some nodes may never be able to transmit. Fixed nodes have various options, such as directional antennas or power levels that can be employed to resolve the issues, but it isn’t as simple with roaming nodes.

A second issue involves different types of interference. The RF layer is one that network engineers tend to ignore when designing the network. They assume all of the nodes are pristine, which would mean the connection is simple and non-congested. In reality, error-free point-to-point links rarely exist with any consistency. That is generally the case with non-wireless networks.

Fig. 1. Hidden nodes. Source: Wikipedia

On the other hand, RF engineers focus on the link and don’t always consider what is happening at the IP level. But there is a lot going on with TCP stack algorithms, TCP window sizing and bandwidth scaling revolving around packet losses and packet delay, as well as UDP and other traffic issues.

Hidden nodes present a unique problem. Because mesh networks are all short-range connections, one node (X) forwarding to another node (Y) can interfere with other nearby nodes on the network and not even be aware of them. The signal strength of the nearby nodes may be too low for nodes A or B to sense and subscribe to the standard 802.11x protocol of “listen then transmit.” They may not hear other traffic well enough to avoid transmitting, especially if the nodes are mobile and signal strengths vary.

This becomes even more of an issue when looking at it from the standards perspective. If everything in the network uses the same protocol that’s great, but it is highly unlikely that such will be the case with the IoE. “Particularly with spontaneous networks, how do you get all types of dissimilar devices to communicate,” asks Rambus’ Woo. “How do you agree on protocols, or a set of protocols and an architecture that will work for both current devices and future devices, especially with something as nascent as the IoE?”

For that reason, mesh networks work well for bursty type of communications, but when the network starts to load up, RF issues such as talkover — an old term for simultaneous transmissions on the same frequency — begin to show up.

One thing that mesh networks have a difficult time with is streaming media, which is the fastest-growing segment of all networks, especially mobile ones. There is a general assumption when one designs a wireline network that all subscribing endpoints will have near-zero percent media errors. With wireless networks, that simply isn’t the case and one way to deal with that is with multicast. As the bandwidth in wireless networks varies, multicast can keep the data coming at a sufficient rate to keep the streaming within acceptable parameters.

Security
Encryption is critical with mesh networks. The idea that data is being routed through and stored on dynamic, often ad hoc, networks of both user and autonomous devices does raises some security concerns.

Fig 2: Security structure. Source: Motorola.

A big concern, in that vein, is with networks that constantly have devices coming and going. That makes the possibility of rogue devices very real. One has to be especially vigilant of someone trying to insert a rogue device into a mesh network, especially the large and open networks that are likely to be part of the IoE. It is totally conceivable that an attacker can insert a rogue device or application, or take control of a device already on the net and use that as an entry point into the network. Therefore, there is a lot of discussion going on about trust within mesh networks.

Another major issue is the lack of management oversight. In traditional enterprise networks, provisioning enables the routing, quality, speed and security the data around the network to be managed centrally. That’s not the case with mesh networks.

At the technology level, the MAC layer is responsible for ensuring the network carries traffic from only authorized stations. This is one strong approach to preventing attacks from rogue stations getting on the network. It does this using a couple of key methods:

• Availability – In the context of wireless networks, availability refers to the network services’ survivability in the face of denial-of-service attacks. There is a self-healing technology in wireless mesh networks that is common to all MANETS, which lets the wireless mesh network route traffic around jammed areas automatically. So if there is an attack happening, at least there is this protocol that can come into effect.
• Authentication – The standard existence of an enrollment mechanism must be integrated. It exists in the 802.11X standard, so it should be workable with mesh nets, as well. It calls for two basic protocols. The first is PSK, which uses pass-phrases or other key material provided to each station in advance. The second is public key, which is the standard, certificate approach that uses use private keys to authenticate each station’s identity.

As the IoE evolves, deployment will uncover the security holes and technology will deal with that. But at least there is a baseline that mesh nets can start with.

Missive
Some people argue that mesh networks will not be a big player in the IoE. Other disagree. Don Dodge, developer advocate at Google, notes that this is a great alternative network for the IoE because “sensor traffic is not meant to have a high quality of service or be high bandwidth. It is not a mesh network for computers. It is meant to connect billions of sensors to the Internet. That is the whole point, and why most longtime network people don’t think in these terms.”

That is an interesting perspective. Wireless mesh networks are sort of the red-headed stepchild of wireless networks, at present. Ad hoc networks present some real management and security issues that will have to be addressed within the IoE universe. They also offer a very good solution to some of the IoE’s challenges. How this will play out is still being debated.

Mesh protocols are interesting and will, certainly, find homes in certain applications. The most challenging issue will be to overcome the performance issues that arise from ad hoc wireless networks, which is the predictability and reliability of planned routed networks with directional links. Wireless mesh networks tend to work well under lightly loaded conditions and bursty traffic, but how they will work in high-traffic situations will have to be determined. So far, there isn’t wide-scale deployment for critical infrastructure needs.

But on the positive side, they are easy to set up and robust. It will be interesting to see if that ease in set up and reliability can scale to high-capacity use cases.