Clearing up the mesh about wireless networking topologies: Part 1 -

Clearing up the mesh about wireless networking topologies: Part 1

Over the past few years, Mesh networks have become more popular,following the trend to create more wireless things. As with othertechnology trends, there have been a plethora of different meshnetworking technologies and architectures.

This series of articles is intended to bring order to the mess of mesh networks. Here in Part 1,network basics will be discussed, focusing on the specifics ofwireless, as well as the criteria for evaluating different wirelessmesh networking technologies.

In Part 2, I will provide an overview of five different mesh relatedtechnologies, including key characteristics, network architecture,strengths and limitations. This information will be aggregated tocreate an evaluation of these different approaches, including when theyshould be applied.

To begin we start with the basics of networking and differenttopologies. It is important to note that network topologies describethe interaction and interconnection of the participants. This means howthey communicate with each other and how they establish paths betweeneach other.

Network topologies are not always what they seem. In the wired days,they generally followed the path of the wires ” very simple. If deviceswere wired in a ring, then the network topology is a ring. The journeyto wireless complicates everything because we all share the same airspace so the path and access method is not always obvious. For example,is a WiFi access point a star topology or a bus?

Some common terminology
Before we go further in looking at these questions, it is important tofirst understand some common terminology which will be used throughoutthis article.

DSSS – Directsequence spread spectrum. This is a method of encoding asignal which distributes information over a wide path of spectrum usinga pseudo random code. Because of the wide spreading, the signal appearsto be noise for those without the spreading code.

FHSS –Frequency hopping spread spectrum. Similar to DSSS, the bigdifference is that it uses a more constrained spreading algorithm andchanges channels as a function of time, theoretically making thetransmission more immune to interference.

TSMP – Timesynchronized mesh protocol. This is a mesh protocol thatuses time slots to allocate spectrum for communication between twonodes. Because time slots differ over pairs, interference is minimizedbecause access to the channel is controlled by timeslot.

AODV – Ad-hoc on-demand distancevector routing algorithm. AODV is a pure on-demand routeacquisition algorithm – nodes that do not lie on active paths do notmaintain any routing information nor participate in any periodicrouting table exchanges. Furthermore, a node does not have to discoverand maintain a route to another node until the two need to communicate.

Cluster-Tree – Region based mesh network routing algorithm. In this algorithm, routesare formed and maintained between clusters of nodes. Route discovery iscompleted and maintained between the clusters ” providing access to thechildren of each cluster.

ISM – Industrial Scientific andMedical band. This describes the frequency bands that can beused license free. Generally we refer to the 2.4 GHz band, but it alsocovers spectrum in 900 MHz, 5.8 GHz, 433 MHz in North America. 2.4 GHzis used worldwide.

IPv6 – Internet Protocol Version 6 . This is the latestversion of the popular IP or Internet Protocol. With Version 6, the IPaddress structure, routing and class of service changes. It is part ofthe TCP/IP suite of protocols sponsored by the IETF.

PAN ID – PersonalArea Network Identifier. This is the term for the networkname assigned to particular personal area network.

CSMA – Carriersense multiple access. This protocol defines the channelaccess technique deployed by Ethernet, WiFi and bus oriented networks.It provides a method for detecting collisions and retransmitting as amethod to acquire a communications channel.

TDMA – Timedivision multiple access. Theprotocol defines the channel access technique used by TSMP and GSMnetworks in which a communications channel is divided into time slots.Each node is allocated a specific time slot for communication.

Network Types
Now that we have a feel for the terminology, we first look at differentnetwork topologies commonly used. Figure1 below shows three such topologies: Star, Bus and Ring.

Figure1. Three typical wireless network topologoes

In the Ring, nodes are connected from one to the next.Communications messages are forwarded around the ring in either aclockwise and/or counter-clockwise fashion.

As a message is forwarded, the node checks to see if the message ismeant for itself, if so, it keeps the message, if not, it forwards iton. It is most common in cabled networks (wire or fiber), but couldconceivably be used in a wireless fashion as well but is not practicalunless being used over long distances.

In the Bus, all nodes share a common communications medium andcontend for using it. Typically this means some kind of CSMA typeapproach. Since a common medium is used, collisions and retransmissionsincrease with traffic loading.

In the wired case, these types of systems are referred to as hubs -which are generally no longer used. In the wireless case, it is morecomplicated because open space is often a shared medium, so even ifrouting is handled in a star, ring or other topology, open space oftenappears as a bus. More on this later as it is one of the fundamentalhurdles in wireless networking.

In the Star, nodes are connected through a master, central node.This central node is responsible for looking at each message andforwarding it out on the proper communications link.

While various star architectures have been used over time, the mostcommonly known in the wired space is the Ethernet Switch. In thewireless space, the WiFi access point is also a familiar example sinceall messages are routed through the access point; however, even thoughmessages are routed through the access point, open space is accessedvia CSMA, a bus type protocol.

Mesh networking is morecomplete
A mesh network employs some level of more complete interconnectionamong nodes. This means that paths are not defined by a specificarchitectural pattern, but rather by the connections themselves.

In the full mesh topology, each node (workstation or other device)is connected directly to each of the others. In the partial meshtopology, some nodes are connected to all the others, but some of thenodes are connected only to those other nodes with which they exchangethe most data. Figure 2 below illustratesa full mesh where each of the five nodes is connected to all theothers.

Figure2. A fully connected wireless mesh network configuration.

Other important thing to note about a mesh is that some or all nodesmay be routers and some or all nodes may be end points. Typically, fullinterconnection is not achieved unless the network is very small.

Full interconnection gets very complex very quickly. In addition,wired mesh networks tend not be practical due to the complexities ofconnecting all the wires.

Figure 3 below illustratesthree different instantiations of mesh networks. The green nodes areend devices, the yellow are routers (which may also be end devices) andthe purple is the network coordinator responsible for allowing joiningand departing from the mesh (more on this later).

Note that one instantiation of a mesh can be a star – a mesh withone router and the rest end points. The Cluster-Tree network is acombination of near full connectivity among routers and end pointshanging off individual routers. The Peer-to-peer mesh generally givesequal rights to all nodes, including routing and end pointfunctionality.

Figure3. Three mesh networking instantiations

What makes wireless networks sodifferent.
While we have discussed that mesh networks aren't really practical forwired networks, it is also important to look at the other differencesbetween wired and wireless domains.

So what else makes wireless different? On the positive side,wireless makes it possible to have more connections since it is notpractical or cost effective to create a full mesh with wires. However,on the negative side, wires are predictable, reliable and wellunderstood.

Wireless forces the sharing of an already noisy, uncontrollablemedium called open space. Hence, while wireless gives us moreflexibility, the uncertainty of wireless drives the need for moreconnection pathsand more complexity.

When evaluating wireless, particularly mesh networks, there are anumber of hard problems that need to be solved, including:

Accessing themedium. Since we all share open space, listening is moreimportant than talking. If everyone talks at once, listening isdifficult. So radios must be good listeners if they are going to have achance to get a word in edge wise, so to speak.

Discoveringroutes. Determining paths in a wireless mesh network isdifficult because the environment is dynamic. In this case, there aretwo choices: Planning the trip in advance, or taking it one step at atime. Often times doing both is best ” this usually involves retracingone's steps and repeating well traveled routes.

Adapting to achanging environment. In a wireless world, paths to nodes candisappear and re-appear. This is due to changing signal conditions ortraffic conditions.

Sleeping andWaking. Once we go wireless, the next step is often to find away to do away with the traditional power cable. This means batteriesand the need for effective power management.

The most common way of handling power management is putting thenodes to sleep when they are not being used. This sounds well and gooduntil it is time to wake up.

How to Compare
Given that we now understand the basics of networks, in particular meshnetworks, the next question that should be addressed is how do wecompare them? For this, we look at criteria of security, reliability,power management, scalability, data movement, and cost.

Security. Thisis as much about the perception of threat as actual threat.Nonetheless, security is easily evaluated by the traditional factorsthat are well understood in the industry.

The first is encryption ” protecting the information itself. Modernencryption wants at least AES128 as an algorithm (128 bit key). Thenext is authentication, which is validating that the users (or nodes)are who they say they are. This is typically handled by key exchange orauthenticated certificate.

Last is authorization, which should be thought of as grantingpermission associated with having the right key or certificate. Beyondthese, there are other factors which are associated with the ease ofdistributing and configuring the authorization and authenticationmechanisms.

Reliability. The best way to think of reliability is the ability of a messageto be delivered to the desired destination on time. If the messagealways arrives at the destination when expected, the network is veryreliable.

Secondly, we want the message to arrive at the destination, even ifit is a bit late. The components for evaluating reliability forwireless mesh networks have to do with the following:

1) Frequency agility: thisis detecting and adapting the network around potential interference.

2) Message loss potential:  this is the question as to whether messages get lost in the shuffle.With all the re-routing and different paths, the network must be verycareful to ensure messages don't get lost and that duplicate messagesfollowing different routes get discarded.

3) Adaptability: This isbest described as the network's ability for changing the routing toaccommodate for nodes disappearing while still preventing lostmessages. This is most effective if done quickly.

4) Single points of failure: Are there any single points of failure, what is the risk of themfailing and how is recovery handled?

Power Management. Themost frequent question asked when discussing wireless sensor networksis how long will my batteries last? As soon as the cord is cut,everyone wants to still keep maintenance low.

Viewed in the context of the network architecture, power managementis analyzed in terms of end nodes, router nodes and networkcoordinators. It is most important to have low powered, power efficientend nodes because they are most likely to be far from traditional wiredpower sources.

The routers are second. Battery powered routers, or routers thatsleep, extend the flexibility of the architecture are necessary.Finally, the coordinator is usually powered. Now, in the context ofnodes that can sleep, we then look at their average power consumption.

his is best assessed by looking at the combination of how they wakeup, how frequently they wake up, total transmitting time and totallistening time. Since the most power is consumed when radios transmit,it is important to keep this to a minimum.

Scalability. How big can the network get before it fails, at least on a practicallevel? All the networks have large physical limits in the 10s ofthousands, but the practical design of the network is always muchsmaller.

This is because scalability is related both to reliabilitymechanisms and nature of the application. If a network never has anyproblems which cause rerouting, then network routing tables will neverchange, meaning cached routes will always work and there will be fewretransmissions or reroutes because of failures. The end result is avery stable network that can be very large.

The other aspect concerns the type and volume of data. This dataflow can be placed into three categories: Dribble Data, Bursty Data andStreaming Data ” and they mean just like they sound.

Dribble data is periodic, infrequent and slow, while streaming datais constant, etc. A network can be very large if the traffic is dribbledata because the flow follows consistent patterns, with plenty ofbandwidth. Sleeping networks do well with dribble data, but scalepoorly with streaming data.

DataMovement. Now we look in more detail at data flow, notfor network scalability purposes, but for raw capacity. There is aclassic trade-off in needs: Does the application require lots of datawith low latency or does it require dribble data with long,non-deterministic latency?

As such, in evaluating networks for data movement, a combination ofthe following five variables needs to be considered: data rate,latency, packet size, fragmentation and range.

 Cost. Cost ismeasured by the individual unit cost as well as the cost to maintainthe network. In this context, maintenance is often difficult toquantify and deployment cost is often forgotten.

It is easiest to quantify those variables that are most perceptible,namely the actual purchase cost of a transceiver system per node. Thisbecomes a bit more complicated when trading off the number of batterypowered or sleeping nodes.

For example, assume all end points need to be sleeping end pointsand a point to multipoint system is not practical due to range. Then anetwork that does not have sleeping routers will need to deploy poweredrouters in addition to the end points as compared to a network that hassleeping routers.

Hence, even if all radios are the same cost, more radios are neededin the powered router system. However, if power is available, then itbecomes a non issue. So, the cheapest radio may not be the best for theapplication. The other point is that the cost of the radio tends to belooked at related to the cost of the device to which it is connected.

Next in Part 2: ComparingZigbee and alternative wirlesss mesh technologies.

Joel K. Young is senior vice president and Chief TechnicalOfficer at Digi International.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.