You are on page 1of 9

Scalability Of WiMax Ad-hoc Mesh Networks

Tycho Bom
t.m.bom@cs.utwente.nl
ABSTRACT

1.1 Research questions

This paper discusses the scalability of WiMax ad-hoc mesh


networks. First an introduction to ad-hoc networks and mesh
networks is given. Different types of mesh networks are
presented and after that, a list of current issues with mesh
networks is presented. To all the issues listed, solutions are
given. When the solutions are described, they are compared
with each other based on certain criteria related to scalability.
Finally the conclusion is drawn that WiMax ad-hoc mesh
networks are very scalable, if the correct solutions are applied.

The question is, how many devices can you connect to a mesh
network? Is the network performance still reasonable, if you
connect a lot of devices? What is a lot of in this case? Can
you connect 10 devices while still enjoying reasonable
performance or is the communication between 100 devices also
possible? These questions can be merged into one main
research question:
How scalable is a WiMax ad-hoc mesh network?

Keywords

The main research question is divided in several sub questions


to split up the problem. The sub questions are as follows.

WiMax, Ad-hoc, wireless, mesh, network, scalability

What different types of ad-hoc mesh networks does


WiMax support?

1. INTRODUCTION

What are the open issues in a WiMax ad-hoc mesh


network?

What are the requirements and criteria that should be


supported by a scalable WiMax ad-hoc mesh network?

How do the WiMax ad-hoc mesh network solutions fulfill


the requirements of a scalable network?

The last couple of years the internet is more and more going
mobile. Almost every laptop is equipped with Wi-Fi (IEEE
802.11g) [IEE03][DVK07] and a lot of cell phones can connect
to the internet via GPRS or UMTS. Like connections at home,
the mobile connections are also getting faster. The next step in
mobile access technologies is Worldwide Interoperability for
Microwave Access, WiMax (IEEE 802.16e)[IEE07][Wim07].
This promising technology can be implemented in a lot of
different devices. Not only phones, PDAs and laptops can use it
in the future, but for example also security cameras or other
household devices can take advantage of the technology. This
means that the number of devices that can be connected to
WiMax is a lot larger than the number of devices that currently
use Wi-Fi or UMTS. Devices of both standards can use WiMax
in the future [She05].
In WiMax networks you do not always need a central access
point to connect all the devices. Besides an infrastructure mode,
WiMax also supports the mesh mode. This means that all clients
together form the ad-hoc network and that every client also
functions as a router and a pass-through for other connected
devices. This extends the network coverage and can offer an
economical advantage [Tal07].
Although a lot of research on WiMax has been done, there are
still some open issues that need to be solved. Not every solution
to those issues scales very well and if WiMax is to be used
widely, it has to be very scalable. There are papers available that
describe a single solution, but more interesting is the
combination of several solutions and the scalability of those
solutions. This paper will give insight in the proposed solutions
for certain issues, compare the solutions based on certain
criteria and draw a conclusion on the scalability of those
solutions.
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or
distributed for profit or commercial advantage and that copies bear this notice and
the full citation on the first page. To copy otherwise, or republish, to post on servers
or to redistribute to lists, requires prior specific permission.

8thTwente Student Conference on IT, Enschede, January 25 th, 2008


Copyright 2008, University of Twente, Faculty of Electrical Engineering,
Mathematics and Computer Science

These sub questions will be answered in different sections of


the paper and after that, the main question will be answered.

1.2 Research method


This paper uses information from existing literature. In the
literature several solutions to open issues are proposed. These
solutions are explained, compared and finally a conclusion on
the scalability of those solutions is drawn. The information
about the solutions is retrieved from the papers itself and to
compare those solutions, certain criteria are derived from other
papers. The conclusion on the scalability is drawn based on the
criteria.

1.3 Structure of the paper


The introduction to the problem has already been done in
section 1. The rest of the paper will answer the sub questions
and finally the main question. Before solutions can be provided,
first an introduction to ad-hoc and mesh networks is given in
section 2. In section 3 a list of current open issues in WiMax
ad-hoc mesh networks is given. Section 4 gives a list of criteria
for the comparison of the presented solutions. This list can be
useful while reading the solutions. The issues from section 3
will be solved using the solutions provided in section 5. In
particular, section 5 provides one or more solutions to each
issue. In section 6 the given solutions to each issue are
compared using the criteria from section 4. Section 7 will give
answers to all sub questions and finally the main research
question. Furthermore, section 7 also provides some ideas for
future work.

2. WIMAX AD-HOC MESH NETWORKS


In this section an introduction to WiMax ad-hoc mesh networks
will be given. This information will be useful for understanding
the rest of the paper.

2.1 Ad-hoc networks

2.2 Types of mesh networks

Basically there are two possible network structures in wireless


networks. The most used structure is the infrastructure mode.
The other mode is the so called ad-hoc mode. In networks based
on the infrastructure mode, also called Point-to-multipoint
mode (PMP) [KYH07], there are typically one or more access
points to which all clients connect to. All clients share the
available bandwidth of the access point they are connected to
[KYH07]. In Wi-Fi and WiMax networks, these access points
function as a gateway or router to the network. In most cases the
access point provides access to the internet or a company
network. Networks based on infrastructure mode are very much
like Ethernet networks. In Ethernet every client connects to a
hub or switch. In wireless networks these cables are replaced by
radio signals. Small infrastructure networks are quite easy to set
up and configure, but if you want to extend the network, it can
become very difficult to maintain the whole network. Also,
infrastructure networks are mostly on a fixed location and
cannot be moved easily because most of the access points need
a connection to the wired network.

In mesh mode, there is no central server or access point. Most


clients also function as a relaying router to other clients. In
mesh mode there are different kinds of stations. One can
distinguish, see e.g., [AWW05], mesh clients, mesh routers and
base stations.

Ad-hoc networks do not need central access points. All clients


form a network together and connect directly to each other. The
main advantages of this type of network are the little
configuration needed and the ease with which the networks can
be moved. Wi-Fi (IEEE802.11) already supports the ad-hoc
operation mode.
The only parameter that needs to be
configured, is use of the same network name. After this
procedure a connection can be performed between two or more
devices. One of the main problems however, is that clients that
are not in direct range of each other cannot connect to each
other. In the infrastructure operational mode the range of the
network is enlarged by using access points, which are operating
as gateways to the Internet. However, also in this operational
mode the range of the network is limited by the range of the
weakest device. WiMax solves this problem by supporting
multi-hop connections. This means that if client A is not in
range of client B, it might connect to client B via client C. This
makes it possible to create a network that is theoretically as
large as the internet. Unfortunately it is not easy to create large
ad-hoc networks. A lot of routing and scalability issues arise. In
WiMax the ad-hoc mode with multihop connections is called
mesh mode.

Mesh routers are the most important devices in the mesh


network. They can connect to other mesh routers and to mesh
clients. Because WiMax supports multihop connections, in this
way the whole network can be linked together. Mesh routers
form some sort of cloud of connected mesh routers with
connections to as many other mesh routers as possible. Mesh
routers have the ability to route all the traffic to the correct
destination using multihop connections, like in wired
connections.
Mesh clients are devices like laptops, PDAs or cell phones that
do not have the ability to function as a mesh router. However, it
is possible that also these devices will be equipped with relay
functionality. Mesh clients connect to mesh routers, which
function as a gateway for the client and provide access to the
internet or some other network.
Base stations are not always necessary, but they can provide
access to devices that do no support WiMax. The base stations
connect to mesh routers and can function as a gateway in
infrastructure mode to for example Wi-Fi enabled devices.
These base stations use the mesh network as backhaul.
Backhaul is the infrastructure that is used to connect base
stations with each other.
The big advantages of a network in mesh mode are the low
costs and the flexibility. It can be deployed very quickly and
requires no or little configuration. Even if some clients or mesh
routers are removed, the network is capable of reorganizing
itself and maintaining a connection, provided there is
redundancy in the connections to other stations.
The following subsections give an overview of the different
architectures used in mesh mode, see also [AWW05].

2.2.1 Backbone architecture


The backbone architecture combines different technologies.
Mesh routers function as a backbone for base stations. Clients
do not directly connect to the mesh routers, but connect to base

Figure 2.2.1 Backbone architecture (copied from [AWW05])

Figure 2.2.2 Client architecture (copied from [AWW05])

stations, which are connected to mesh routers. The mesh routers


replace the traditional wires to connect the base stations with
each other and the internet or external network. Figure 2.2.1
gives an illustration of the backbone architecture.

be solved.
1.

Radio interference, see [FY06][HGI+05]. Because relay


stations or clients may want to send data at the same time,
the available radio frequencies have to be shared.

2.2.2 Client architecture

2.

Movement of clients, see [GSS+06]. If clients are moving,


the connection has to do a handover to another access
point or relay station.

3.

Quality of service, see [HGI+05]. In WiMax mesh


networks this is difficult to achieve. Because clients use
multihop connections, it is difficult to manage the
bandwidth from one single access point.

4.

Load balancing, see [Tal07]. Because you do not want to


overload some parts of the network, while other parts are
not used, you need to balance the load.

The client architecture is the most simple form of mesh


networks. There are no different roles for devices to play. Each
connected device functions as a relay station to other devices.
Each device performs routing and configuration tasks and
provides applications to end users. Typical clients are laptops,
cell phones and PDAs. The requirements on clients in this
architecture are increased compared to infrastructure networks,
because each client has to perform additional tasks. Figure 2.2.2
gives an illustration of the client architecture.

2.2.3 Hybrid architecture


The hybrid architecture combines the backbone architecture and
the client architecture. Clients can connect both to other clients
as well as to mesh routers or base stations. Base stations can
provide the clients with additional technologies such as Wi-Fi
or Bluetooth. Figure 2.2.3 gives an illustration of the hybrid
architecture.

3. WIMAX AD-HOC MESH ISSUES


When creating ad-hoc WiMax mesh networks, a lot of issues
arise. Not only performance issues need to be addressed, but a
lot more issues can limit the scalability. Below a list of issues to

4. CRITERIA FOR COMPARISON


The solutions to the issues found, will be compared with each
other. Below a set of criteria is listed. These criteria will be used
to examine the different solutions..
1.

The number of users that can be supported by the solution.

2.

Coverage area, see [NC06]. There are not only routing


issues and software related issues. Also the range of
WiMax devices can be extended to enlarge the coverage
area of the network.

Figure 2.2.3 Hybrid architecture (copied from [AWW05])

3.

4.

Configuration complexity in relation to scalability, see


[ZTS+07]. Some solutions introduce a topological and
hierarchical architecture for ad-hoc networks. This means
more configuration and less flexibility.
The number of mesh hops used while still providing
reasonable performance.

5. SOLUTIONS
In this section each issue described in Section 3, will be further
discussed and for each issue one or more solutions will be
presented. This will be presented in the same order as presented
in section 3.

5.1 Radio interference


In WiMax multiple-access radio interference is a major limiting
factor for scalability and speed. [HGI+05] proposes a solution
to this problem with an interference-aware cross-layer design.
They use a tree-based routing framework, with scheduling to
increase the usage of available radio frequencies. A tree-based
routing framework means that each node is connected to one
parent node and that each node can have several child nodes.
The solution achieves a significant throughput enhancement
over the basic IEEE 802.16 mesh network. The paper
introduces a new network layer that relies on the load demand
of the Application Layer, the interference information of the
Physical Layer and scheduling and route selection of the Data
Link Layer. The paper introduces a blocking metric for a
multihop route. This means that in the routing decision, the
interference with other nodes on that route is considered. If a
node wants to send data to another node, all other nodes in the
range are blocked for sending. In this way collisions are
avoided. The routing algorithm proposed, selects the route that
blocks the least other nodes.
[HGI+05] also proposes interference-aware scheduling. The
scheduling algorithm seeks to maximize the number of
concurrent transmissions, without creating interference. When a
node wants to connect to the mesh network, it has to determine
its sponsoring node to connect to. The new node selects the
sponsoring node based on the blocking metric. With the
information from the network entry and the initialization
process, the scheduling algorithm can iteratively determine a set
of active links at some time. With this information it can
allocate the traffic demand for as much concurrent
transmissions as possible.
[FY06] also proposes a solution to the limited radio resources
available. It does not create a technical solution, but it proposes
a way to calculate the optimal number of relay stations in an
area. Although relay stations extend the coverage of the
network, too many relay stations that can communicate with
each other will make the use of radio resources inefficient. The
paper uses mathematical formulas to calculate the use of radio
resources used for backhaul and for last hop access. The
conclusion drawn is that radio resources can be used optimal
when you deploy as less relay stations as possible, while
covering the whole desired area. The paper also concludes that
one should prefer base stations with a low coverage radius over
base stations with a high coverage radius. In this way, the
minimal radio resources for backhaul are wasted and the fastest
connections to end users can be provided. The disadvantage of
using base stations with a small coverage range, is the higher
costs. On the other hand, implementing base stations with a
high coverage range is cheaper, but will provide worse speed to
the end users.

5.2 Movement of clients


WiMax is a wireless technology, so a lot of clients will be
movable devices. WiMax can cover a large area, but when a
device goes out of range of one mesh router, it has to connect to
another mesh router. This means that the network has to be able
to reroute the traffic to the client.
[GSS+06] proposes a solution to the problem of searching for
new gateways. The paper describes the problem in the case of
providing internet in public transport, like busses. Each bus
provides internet access to its customers. All the buses together
form one mesh network. Because the busses are moving, they
have to search very often for another gateway to connect to. The
paper assumes that the routes the busses take are known in
advance. This information is entered in the routing protocols
and it shows that de delay of packets travelling from the
gateway to the bus is decreased a lot. This means that their
solution works, but of course it only works when the route of
the movement is known in advance. So the solution is only
suitable for very specific situations.

Figure 5.2.1 Network topology (copied from [CN04])


[CN04] also proposes a solution to the issue of moving clients.
It introduces the hierarchical ad hoc network. This means that
although it is an ad hoc network, it does use some hierarchy to
provide connectivity to the internet or some other network. The
access points, see figure 5.2.1, are interconnected by a high
speed backbone. The network has several sub networks, that are
formed based on the geographical location of nodes. The idea is
that there are some central access points with moving nodes,
called runners, to support the nodes that want to connect. The
central access points are connected to the internet or some other
network. The runners are support nodes for end users and
connect to the access point and other runners. The runners are
constantly moving in for example cars. These nodes provide
access to end users.
In figure 5.2.1 each city has an access point, called an access
port. On the highways and in the cities, the support runners are
travelling. This creates a very dynamic network, similar to the
solution provided by [GSS+06].
Figure 5.2.2 shows the protocol stack that is proposed in
[CN04]. The paper introduces several new layers to be able to
provide connectivity to moving hosts. The hierarchical layer
provides communication between the runners and the access
ports. The adaptation layer checks if the network is changing.

Figure 5.3.1 Network topology (copied from [CTS03])

Figure 5.2.2 Proposed protocol stack (copied from[CN04])


Changes can be caused by nodes that move out of range or
nodes that come in range. Based on that information the
adaptation layer decides if more support nodes are needed or
not. It will always try to enable the optimal number of support
nodes. This is of course not always possible because runners
might not be in the neighborhood.
Below the adaptation layer are the protocols that are used in the
runners or support nodes. The protocols detect whether there
are nodes in range that want to access the network and they
keep track of the movement of those nodes.
Sending of some data from A to B works as follows:

Host A gets notified by the sensor sub protocol when a


mobile host of the support (a runner) is in its
communication range. Host A then gives the message to
this runner.

When the runner comes in the communication range of an


access port, it transmits the message from A to that access
port.

The access port transmits this message over its own


backbone network to the access port of host B, the
receiver.

The access port of host B will forward the message to a


runner, which will in turn give the message to host B.

What you see in this approach, is that the connection is not


always real-time, but if there are enough runners available, realtime connections can be made.

5.3 Quality of service


A lot of networks support Quality of Service (QoS) nowadays.
Because mesh networks use multihop connections, it is difficult
to guarantee certain bandwidth and delay to end users.
[CTS03] proposes an on demand, link state, multipath routing
protocol. The protocol uses tickets to ask for guaranteed
bandwidth and uses multipath routing to fulfill the bandwidth
request. It works as follows. Suppose sender S, see figure 5.3.1,
wants to send some data to destination D with guaranteed
Quality of Service. Then it first sends a ticket to reserve
bandwidth. If all hops from S to D can reserve enough
bandwidth, the request is successful immediately. If one or
more hops cannot reserve the requested bandwidth, the ticket is
split in multiple smaller tickets. These tickets follow distinct
routes and can together provide the bandwidth requested.
In addition to multipath routing, the protocol also uses the linkstate information to determine the routing path. The link-state
information is collected during the QoS multipath searching. A
long packet records the information of all hops from source to
destination. Each host knows the available bandwidth to its

neighbors. When it needs a route to destination D, it sends out


multiple Route REQuest (RREQ) packets which carry the path
history and link-state information. D will receive multiple
RREQ packets from different paths and can combine that
information. Using that information, it can construct the optimal
route from source to destination. When it has determined the
best route, it sends back a confirmation packet to the source,
confirming the route and reserving the bandwidth. The big
advantage of the routing protocol is, that it uses the information
of the whole path instead of hop-by-hop routing. In this way
multipath routes can be valued in terms of QoS more accurately.
[HV05] proposes a total QoS solution. They have designed
algorithms to provide end-to-end Quality of Service guaranteed
for both real and non real-time individual traffic flows. To
support the Quality of Service it also provides an admission
control policy. [HV05] assumes that IEEE 802.16d is used, so
all stations are on a fixed location. This could be a problem, but
there are a lot of situations in for example home connections,
where stations do not move. In those situations, this paper can
provide a good solution.
To be able to reserve bandwidth along the path, they propose a
routing protocol that chooses a fixed route. This means that
when node A sends something to node B, all packets will travel
along the same path. They only make an exception when a link
breaks down or the statistics of the links on some route degrade
dramatically. The routing protocol does not split the traffic and
does not send it over two different links reaching the same
node. This causes unnecessary delays to reorder the packets.
The result of the routing protocol is a tree based scheme. The
routing protocol uses shortest path routing to minimize the
average work needed and transmission time. The root of the tree
is a base station. For an example of shortest path routing, see
figure 5.3.2. The left side of the figure shows the network with
all nodes and physical connections to other nodes. The number
next to each connection denotes the cost of that link. The cost
can be based on link speed and delay. The right side of the
figure shows the routing scheme. In this scheme every node
uses the links with the lowest costs to reach other nodes. Also
you see some sort of tree in the scheme. This means that each
node is only connected to one higher node.
The paper distinguishes real time and non-real time traffic. Real
time traffic is for example IP telephony or video streaming.
Non-real-time traffic is traffic like web browsing or file transfer.

Figure 5.3.2 Shortest path routing (copied from [HV05])

To ensure QoS in real-time traffic, they first propose a


mechanism to make sure the delay will be small. Data is always
send in frames with some specified length. After sending each
frame, the packets that could not be transmitted in that frame are
dropped immediately. This could mean that data might be lost,
while it could successively be buffered. This is not true, because
if the delay from sender to receiver is too high, it will be
dropped at the receiver anyway. So if it is known in advance
that the packet will not be received on time, it might be better
not to send the packet at all, because it will cause useless load
on the communication network. The paper assumes that the
maximum packet loss requirement should be 2% for real-time
applications to work correctly.
Now the guarantee on the delay is set. Now it has to be made
sure that less than 2% of the packets will be dropped. [HV05]
uses a central base station to coordinate the QoS requirements
and the traffic. This base station is the root of the routing tree.
Mathematical formulas are used to calculate the number of slots
needed for each node. The base station then can give that many
slots to each node in a given frame so the QoS constraints are
met. For the complete solution, see [HV05]. More interesting
are the results of the performed experiments. The simulations
prove that the solution really works. In [HV05] is shown that
the probability that a packet is dropped, is 0,0002%, being
much better than the to be fulfilled requirement. Also the delays
are much lower than they stated as reasonable. According to the
paper a delay of 150ms for real-time applications is acceptable.
The results of the simulations in [HV05] show that the delays
are varying between 25ms and 27ms.
[HV05] also proposes a QoS scheme for TCP connections.
Although this operates differently, it is also coordinated from
the base station. For details, see [HV05]. More important is that
the paper proposes an admission control mechanism. Each node
that wants to setup a connection with QoS requirements, sends a
request to the base station for that connection. The base station
then determines if it can fulfill the request or not. If it can, the
bandwidth along the path is reserved for that node. If it cannot
fulfill the request, the request is rejected. The node can place a
request for both TCP and UDP (real-time) traffic.

5.4 Load balancing


If all clients and routers can determine their own routes, it is
possible that some parts of the network will be overloaded,
while other parts are unused. This is not desirable, so some load
balancing has to be done.
[Tal07] provides a load balancing algorithm. The basic idea is,
that it divides the mesh network in several disjoint clusters.
Each node is connected to one cluster. Within each cluster there
is one gateway connected to the wired network and serving the
other nodes. In each cluster the gateway is the root of a
spanning tree. The spanning tree is used for traffic aggregation
and forwarding. For QoS constraints each cluster has a
maximum size of nodes and an upper bound on the cluster
radius. The cluster radius is counted as the maximum number of
hops. There is also an upper bound on traffic between relays.
Gateways can be placed to divide the network in the minimum
number of clusters, satisfying the QoS constraints.
When dividing the network in several disjoint clusters, it shows
that the load balancing factor is much higher than in normal
situations. It does not really matter how many clusters you add,
so the network also becomes a lot more scalable.
[JFZ+05] proposes an algorithm for using concurrent
transmissions in multihop connections. In this way the end-to-

Figure 5.4.1 (copied from [JFZ+05])


end throughput is improved and the traffic load is better
balanced. It also takes radio interference into account when
determining the routing paths. The solutions works as follows.
[JFZ+05] assumes that each network starts with one base
station, see figure 5.4.1. Each client that wants to connect to the
network, has to select a father node. This father node is selected
based on minimal interference. After the node entered the
network, the interference may have changed, so the routing tree
is changed accordingly..
When a node wants to send data, it broadcasts a request. This
request is picked up by its father and forwarded upwards the
routing tree. In the same way other nodes, that do not interfere
with this node, can also send data. So in the system concurrent
transmissions are possible. This balances the traffic in the
network and prevents it from overloading some parts of it.

6. COMPARISON
The provided solutions for each issue are compared below.
They are compared using the criteria listed in section 4. After
the solutions to each issue are compared, a table will be given to
summarize the comparison. A + indicates that the solution has a
positive effect on the criterion and a indicates that is has a
negative effect on the criterion. If a cell is empty, the effect is
neither positive or negative.

6.1 Radio interference solutions


[HGI+05] and [FY06] both solve the problem with radio
interference, but in a different way. When we first look at the
number of users that both solutions provide, [HGI+05] scores a
lot better than [FY06]. [HGI+05] does not limit the number of
users, but it creates a technical solution by blocking other nodes
that could interfere with the sending node. This means that they
can be connected, but they have to wait sometimes. So if a lot
of users are connected and want to send data, the performance
will decrease. [FY06] does limit the number of users by
calculating the maximum users or nodes in an area. So they are
not only blocked for the time other nodes are sending, but they
cannot even connect. By totally blocking users, [FY06] will
likely provide a better performance, but will support less users
than [HGI+05]. When implementing a network, you will have
to choose if you want to support a lot of users or that you want
to make sure the provided performance is high.
On the criteria of the coverage area, the solutions both have
their advantages. In principle [HGI+05] should provide better
coverage than [FY06] because it supports more nodes and this
will result in a more dense coverage. However, [FY06] allows
all useful nodes to connect. This means that it will try to
maximize the network coverage, but it will block nodes that
interfere or do not extend the network coverage. So the
coverage of [HGI+05] will be better, but the interference with
[FY06] will be less because it simply allows less nodes.
The configuration of [FY06] is more complex than the
configuration of [HGI+05]. The first will need some

management to allow nodes to connect or not. The latter will


manage itself by using the blocking metric in the routing
protocol and deciding which nodes are blocked when some
node is sending.
Regarding the number of mesh hops the solution can support
while providing reasonable performance, [FY06] scores better
than [HGI+05]. Because [HGI+05] does not limit the number of
nodes in some area, the network will become very slow when a
lot of nodes interfere and have to wait for other nodes to stop
sending. So the data can travel along a lot of nodes, but it will
become very slow. [FY06] does not have this problem because
it limits the number of nodes, there will be less interfering
nodes and nodes will have to wait less. This means that it can
support more hops in a connection while providing reasonable
performance.
Table of the comparison in 6.1
[HGI+05]

[FY06]

Number of users

Network coverage

Configuration complexity

Number of hops supported

As can be seen in the table of comparison 6.1, [HGI+05] scores


better on most criteria than [FY06]. This means that the first is a
better solution to the issue of radio interference in ad hoc mesh
networks than the latter. However, [FY06] can be useful in
managed networks.

6.2 Movement of clients

Overall [CN04] scores better on the criteria used than


[GSS+06], but it can depend on the situation which solution is
the best. [GSS+06] can only be used when the movement of
nodes is known in advance, while [CN04] needs a network of
supporting nodes.

6.3 Quality of service


Both [CTS03] and [HV05] propose a solution to provide
Quality of Service in WiMax mesh networks. Considering the
number of users the solutions support, [CTS03] scores best.
[HV05] uses a central base station to coordinate the QoS
requirements, so that station will have a limit on the number of
clients. [CTS03] is a solution that is not based on a single
coordination point. This means that [CTS03] can be more
scalable and it can support more users. However, the difference
in the number of users will only be visible when connecting a
very large number of users.
These solutions do not influence network coverage. They only
optimize and route the traffic over existing links.
At the configuration complexity, [CTS03] is much better than
[HV05]. This is because in [CTS03] each node arranges the
QoS for itself, while in [HV05] a central base station is needed.
This base station needs configuration, so it is more complex to
implement and less scalable.
When the number of hops supported is compared, [CTS03] will
perform better than [HV05]. This is because [HV05] only uses
one path to the destination and does not split up the data over
several links, reaching the same destination. So [CTS03] will
take more advantage of the available bandwidth and thus can
support more hops.
Table of the comparison in 6.3

[GSS+06] and [CN04] both provide a solution to the issue of


moving clients. Based on the available information in the
papers, it is not possible to conclude which of the two solutions
supports the most number of users. Both solutions should be
able to handle lots of users.
In both solutions the access points for the end users are moving.
So it really depends on the number of access points and their
location how good the network coverage will be. Because
[CN04] uses many more access points than [GSS+06] it is
likely that [CN04] will provide better network coverage than
[GSS+06].
In configuration complexity [GSS+06] is much lower than
[CN04] because [CN04] uses a hierarchy and needs central
access points. [GSS+06] on the other hand does require
knowledge of the bus schedules, but that is all. The busses do
not need wired links between them or another connection than
to other busses. So [GSS+06] is better scalable than [CN04].
If we take a look at the number of hops the solutions support,
[CN04] scores better than [GSS+06]. The first uses a
sophisticated network of supporting nodes that can
communicate with other supporting nodes and the central access
points. [GSS+06] only supports the backhaul between the
busses and a one-hop connection to end users.
Table of the comparison in 6.2
[GSS+06]

[CN04]

Number of users
Network coverage

Configuration complexity

Number of hops supported

[CTS03]

[HV05]

Configuration complexity

Number of hops supported

Number of users
Network coverage

To conclude, [CTS03] is a better solution than [HV05] for


providing Quality of Service in WiMax ad hoc mesh networks.
At all criteria [CTS03] scores best.

6.4 Load balancing


[Tal07] and [JFZ+05] both provide a mechanism for load
balancing. [JFZ+05] will be able to support more users than
[Tal07] because it does not put constraints on the number of
connected clients. [Tal07] does put a limit on the number of
clients, but it can form multiple cells.
Both solutions do not directly influence the network coverage.
[Tal07] is more complex to configure than [JFZ+05] because it
needs the network to be split in different cells. [JFZ+05] does
not need different cells and will be able to set up itself.
The number of hops supported is larger in [JFZ+05] than in
[Tal07]. This is because [Tal07] puts an upper limit on the
radius of the mesh network.
To conclude, [JFZ+05] scores better than [Tal07] on the criteria
mentioned. This means that in ad hoc networks, [JFZ+05] is a
better solution to the issue of load balancing than [Tal07].
However, the solution provided in [Tal07] might give better
performance in structured networks, but that is not in the scope
of this paper.

[FY06]

Table of the comparison in 6.4


[Tal07]

[JFZ+05]

Configuration complexity

Number of hops supported

Number of users
Network coverage

7. CONCLUSIONS AND FUTURE WORK


This paper showed several aspects of WiMax. First we saw that
WiMax supports different types of mesh networks. These types
are the backbone architecture, the client architecture and the
combination of those two, the hybrid architecture. The open
issues for WiMax apply to all of these architectures, but some
solutions can only be applied to one of the architectures. For all
the issues discussed in this paper, solutions are available. To
keep the network scalable, it is important that the configuration
complexity also scales well. In this paper several solutions with
a low configuration complexity are discussed, so these solutions
should be chosen over solutions with high configuration
complexity. When applying a combination of the solutions
provided, WiMax mesh networks are very scalable and capable
of connecting hundreds of devices. Especially a combination of
[HGI+05], [CTS03] and [JFZ+05] would be interesting to
implement and should give good results. These solutions
combine the best of all solutions. Which solution to the
movement of clients you should choose, depends on the
situation. Both [GSS+06] and [CN04] can be useful.
Future work should consist of testing the combination of
solutions and standardizing the solutions. Without
standardization, none of the solutions will become a success. To
make the solutions a success, the costs of the implementation
should be taken into account too. Also, the effects of solutions
on power usage should be further researched. Because mesh
clients have to perform additional tasks over just accessing the
network, this can have negative influence on battery power of
mobile devices. Power usage is not taken into account in this
paper, because not enough information about power
consumption is available.

ACKNOWLEDGEMENTS
I would like to thank Georgios Karagiannis and my fellow
students for their useful reviews of this paper and their
comments.

REFERENCES
[AWW05] Akyildiz, I.F., Wang, X. and Wang, W. Wireless
mesh networks: a survey. Computer Networks, 47
(4). 445-487. (2005)
[CN04]

Chatzigiannakis I., Nikoletseas S. Design and


analysis of an efficient communication strategy for
hierarchical and highly changing ad-hoc mobile
networks Mob. Netw. Appl., 9 (4) (2004) 319-332

[CTS03]

Chen, Y.-S., Tseng, Y.-C., Sheu, J.-P. and Kuo, P.H. An on-demand, link-state, multi-path QoS
routing in a wireless mobile ad-hoc network.
Computer Communications, 27 (1).9 (2004) 27-40.

[DVK07] Dekleva S., Shim, J.P., Varshney, U. and Knoerzer,


G. Evolution and emerging issues in mobile
wireless networks. Commun. ACM, 50 (6). (2007)
38-43.

Florea, A. and Yanikomeroglu, H., On the


scalability of relay based wireless networks. in
Wireless Communications and Networking
Conference, 2006. WCNC 2006. IEEE, (2006), 242245.

[GSS+06] Gunadi, S., Samuel, I., Salil, S.K., Quan Jun, C. and
Kun-Chan, L. Feasibility study of using mobile
gateways for providing internet connectivity in
public transportation vehicles Proceedings of the
2006 international conference on Wireless
communications and mobile computing, ACM
Press, Vancouver, British Columbia, Canada, 2006.
[HV05]

Harish, S. and Vinod, S. Algorithms for routing and


centralized scheduling to provide QoS in IEEE
802.16 mesh networks Proceedings of the 1st ACM
workshop on Wireless multimedia networking and
performance modeling, ACM Press, Montreal,
Quebec, Canada, 2005.

[HGI+05] Hung-Yu, W., Ganguly, S., Izmailov, R. and Haas,


Z.J., Interference-aware IEEE 802.16 WiMax mesh
networks. in Vehicular Technology Conference,
2005. VTC 2005-Spring. 2005 IEEE 61st, (2005),
3102-3106 Vol. 3105.
[IEE07]

IEEE Standard for Local and Metropolitan Area


Networks Media Access Control (MAC) Bridges
Amendment 5: Bridging of IEEE 802.16. 802.16k2007 (Amendment to IEEE Std 802.1D-2004).
(2007)
http://ieeexplore.ieee.org/xpls/abs_all.jsp?tp=&isnu
mber=33683&arnumber=1603394&punumber=106
76

[IEE03]

IEEE standard for information technologytelecommunications and information exchange


between systems- local and metropolitan area
networks- specific requirements Part II: wireless
LAN medium access control (MAC) and physical
layer (PHY) specifications (2003)
http://ieeexplore.ieee.org/iel5/8601/27242/0121062
4.pdf?tp=&isnumber=27242&arnumber=1210624&
punumber=8601

[JFZ+05] Jian, T., Fuqiang, L., Zhihui, Z. and Zhangxi, L.,


Throughput enhancement in WiMax mesh networks
using concurrent transmission. in Wireless
Communications, Networking and Mobile
Computing, 2005. Proceedings. 2005 International
Conference on, (2005), 871-874.
[KYH07] Kejie, L., Yi, Q. and Hsiao-Hwa, C. WIRELESS
BROADBAND ACCESS: WIMAX AND
BEYOND - A Secure and Service-Oriented
Network Control Framework for WiMAX
Networks. Communications Magazine, IEEE, 45
(5). (2007) 124-130.
[NC06]

Nmeth Z. and Csaba, A.S. Measurements to assist


access network design with fixed WiMAX in urban
environment Proceedings of the 1st international
conference on Access networks, ACM Press,
Athens, Greece, 2006.

[She05]

O'Shea, D., WiMax makes a mesh. Telephony


Magazine, (2005, October 17)
http://telephonyonline.com/mag/telecom_wimax_m
akes_mesh/

[Tal07]

[Wim07]

Talay, A. A Gateway Access-Point Selection


Problem and Traffic Balancing in Wireless Mesh
Networks. in Applications of Evolutinary
Computing, (2007), 161-168.
WiMAX Forum Mobile System Profile Release
1.0 Approved Specification (2007-05-02)

http://www.wimaxforum.org/technology/documents
/wimax_forum_mobile_system_profile_v1_40.pdf
[ZTS+07] S. Zhao, K.T., I. Seskar and D. Raychaudhuri.
Routing Protocols for Self-Organizing Hierarchical
Ad-Hoc Wireless Networks WINLAB, Rutgers
University, Piscataway, (2007), 4.

You might also like