Why Are There So Many Routing Protocols for Wireless Mesh Networks?

1 downloads 0 Views 51KB Size Report
Abstract — Wireless Mesh Networks (WMNs) are a type of radio-based net- work systems which ... For example, municipal wireless broadband rolled- out using ...
Irish Signal and Systems Conference, Dublin, June 28–30, 2006

Why Are There So Many Routing Protocols for Wireless Mesh Networks? Karol Kowalik and Mark Davis Communications Network Research Institute, Dublin Institute of Technology IRELAND E-mail: [email protected] and [email protected] Abstract — Wireless Mesh Networks (WMNs) are a type of radio-based network systems which require minimal configuration and infrastructure. They can be build using relatively low cost radios and inexpensive computing platforms, and consequently appear to be a compelling option for rolling out networks with a low deployment cost. One of the factors which influences the performance of WMNs is the routing protocol used. There are many routing protocols for WMNs, we estimate that there are more than a hundred of them. In this paper we attempt to answer the question why there are so many and if there is a need for such an abundance. Moreover, we elaborate on the possible development of a single routing protocol for WMNs. Keywords — Wireless, Mesh, Routing, 802.11

I

I NTRODUCTION

Wireless Mesh Networks (WMNs) are a type of radiobased network systems which require minimal configuration and infrastructure. Networks which are created using this technology are intended to be decentralized, low cost and resilient. In such networks each node provides identical functionality and communicates with neighbouring nodes which are within range of its radio in order to exchange routing information and to forward data packets. This exchange of routing information allows a node to know where to forward data packets. The forwarding of data allows other nodes to communicate. Each node exchanges information only with its neighbours, thus the coverage of the network depends on the underlying radio technology, number of nodes and their location in relation to one another. There is no infrastructure required to roll-out such a network, therefore the deployment cost can be low. Mesh networks have potential for reliability because they follow a completely decentralised approach. When a single node goes down its neighbours simply find another route around it. However the deployment of mesh technology for wireless communication is not yet ubiquitous. One cannot purchase a mesh node from a computer store, set it up and allow it to join up with a mesh network. Currently there are many research projects1

which offer network connectivity for local communities. Also mesh technology finds many applications in wireless multi-player gaming, building automation, campus connectivity, military communication, municipal networks, etc. The choice of radio technology for WMNs influences the performance of the network and thus the routing protocol needs to be aware of it, and cannot operate in the same way as wired networks which are agnostic about the underlying medium. The physical radio layer determines the capacity of a link and the quality of the communication, while the media access control (MAC) layer coordinates access among the distributed nodes. The network layer and its routing functionality must be tailored to support mobile nodes, dynamic topologies and changing link capacity, which characterises the wireless mobile environment. Researchers have proposed a variety of new routing protocols developed specifically for ad-hoc wireless mesh networks2 . In this paper we will try to answer the question why there are so many of them. In the following sections we will identify the factors which have resulted in such a high number of solutions. Moreover, we consider if there is any promise for a single WMN routing protocol.

2 an 1 for

a list please refer to: http://www.freenetworks.org/

extensive list can be found at: http://en.wikipedia.

org/wiki/Ad hoc protocol list

II

One of the explanations for the variety of routing protocols for WMNs is the lack of a standard which would define its operation. The IEEE 802.11 is a set of WLAN standards that define many aspects of wireless networking. One such aspect is mesh networking which is currently under development by the IEEE 802.11s task group. This task group has issued a call for proposals and received 15 of them in response, from which two have been selected. The standard is expected to be approved by 2008. However, currently there is no accepted standard what leaves the various aspects of the functioning of WMNs open to public debate. Therefore, many research groups working with WMNs and also companies which provide commercial products are developing proprietary WMN solutions. The interoperability among these different approaches is very limited and so it is preferred to use a single vendor’s equipment. That is why many interested parties prefer to wait until the technology matures and a single standard emerges. III

QoS

T HE LACK OF A STANDARD

T HE VARIETY OF APPLICATIONS

Another important reason for the abundance of routing protocols for WMNs is the range of applications for this technology, and these applications can have diverse operational requirements. For example, municipal wireless broadband rolledout using WMNs require: • high scalability – because it needs to provide access to a large number of customers, • security – to enable secure communication over radio technology,

scalability

nodes mobility

storage traffic patterns

WMNs routing protocol processing

power consumption

security interoperability efficient use of medium robustness

Fig. 1: Various elements influencing the design of a routing protocol for WMNs

In summary, routing not only needs to match the radio environment but also it needs to be tailored to a specific application. Consequently, it is difficult to find a single routing protocol which could be applied to all of these applications. IV

T HE VARIETY OF DESIGN METHODS

When embarking on the development of a routing protocol, there is considerable scope for making design choices. Routing protocols can operate in many ways because there are various methods used for paths computation, distribution of routing information, various data structures for storing such information and several strategies for nodes coordination. Through the selection of such elements a routing protocol realises the goal which its designer intended. In order to discover routes between nodes one needs to gather information about the underlying network topology. All nodes, in order to know how to communicate with other nodes, need to gather such routing information and store it locally. There are two approaches for gathering such information:

• quality of service – to ensure user satisfaction, and to allow for interactive multimedia services.

• proactive – which disseminates routing information to other nodes within the network even before it is needed.

While a mesh network constructed for sensors monitoring say weather conditions typically requires:

• reactive – which gathers routing information ondemand when it is required.

• efficient power management – because the energy provided by the batteries is limited,

The reactive method works well in a wireless environment in presence of mobile nodes and continuously changing topology. The availability of bandwidth in 802.11 networks is scare so the on-demand methods can help conserve it. Reactive methods are widely accepted for WMNs, and therefore many routing protocols follow this approach, examples include DSR [1] or AODV [2]. Proactive methods attempt to provide up-to-date state information for all nodes in the network and have been proven to work well for wired networks, but it is recognised that they scale poorly in mobile WMNs. However, there are also methods which limit the scope and frequency of dissemination of such routing information thus resulting in a more robust and scalable proactive routing protocols such as HSLS [3] or OLSR [4].

• minimal storage – because of the limited amount of memory on such devices, • minimal processing – due to their small processing power, • efficient dissemination of status information – to minimise all three above requirements. Other applications such as wireless mesh gaming, building automation or military communication have their own particular requirements. These diverse requirements, which are shown in Figure 1, tend to influence the design of the routing protocol.

WMNs not only need to gather connectivity information, but they also need to store it. There are two main design choices for its storage: • link state database – which consists of a full or partial view of the network topology. • distance vector – which is a single vector representing the cost of reaching other nodes in the network. The first choice requires the use of an algorithm such as Dijkstra’s, while with the second approach one can calculate routes using the Bellman-Ford algorithm. Both algorithms compute least cost paths, using one of the performance metrics as a link cost. In fact there are many performance metrics that can be used to represent the link cost and new ones continue to be created to facilitate more sophisticated path selection criteria. The most popular metrics used as a link cost for WMNs are: • Expected Transmission Count (ETX) [5] – which is the loss rate of broadcast packets between pair of nodes, • Round Trip Time (RTT) – which is the round trip delay between pair of nodes, • Hop Count – which is the number of links between pair of nodes. An evaluation of the performance of routing protocols with different metrics is presented in [6]. In the following we describe some of the routing protocols which were designed for WMNs, showing the variety of methods and design choices. Dynamic Source Routing (DSR) [1] protocol operates on-demand, allowing one to minimise the overhead because it reacts only to changes in the routes that are currently in use. So, it works well for mobile nodes, resulting in small overhead when nodes are stationary and increasing overhead when nodes begin to move. Hazy Sighted Link State (HSLS) [3] routing protocol limits the scope of topology information dissemination in time and space, so nodes which are located far away receive topology information less frequency that these which are close to a node. So, this protocol can efficiently distribute state information thus, thus its applicable to scalable QoS routing protocol for wireless networks. Ad hoc On-Demand Distance Vector (AODV) [2] routing protocol allows nodes to obtain on-demand routes for a new destination. It features messages for reporting link breakages and changes in network topology to other nodes, thus it can take appropriate actions when such events occur. It is a distance vector protocol and to avoid loops which are often associated with this class of routing protocols it uses destination sequence numbers. Thus it effectively disseminates routing information in the mobile environment.

Optimized Link State Routing Protocol (OLSR) [4] is a modification of a typical link state algorithm optimized to meet the requirements of a mobile wireless networks. It minimises overhead associated with dissemination of state information by the use of multipoint relays, which are the only nodes taking part in process of dissemination of state information. The OLSR protocol minimises update overhead by limiting the number of nodes which are allowed to generate state updates. Thus this technique is suitable for large and dense networks. The mechanisms used to gather routing information, store it and compute paths based on metric must be carefully combined to obtain stable routing performance. The choice of these methods depends on the application for such routing schemes. However, one set of such methods cannot fulfill all the requirements of a wide variety of applications. V

T HE BANDWIDTH PROBLEM

The availability of bandwidth in 802.11 networks is scare and the methods which aim to increase the throughput per node will also influence the way the routing protocols operate. The 802.11 standard defines two methods for accessing the medium: distributed coordination function (DCF) and point coordination function (PCF). DCF is mandatory and is based on the CSMA/CA (carrier sense multiple access with collision avoidance) protocol. With DCF, stations must compete for access to the medium. When the medium becomes busy stations must defer their transmission until it becomes idle again. In multi-hop mesh networks packets need to compete for medium access at each hop, thus the throughput is reduced at each hop. Gupta and Kumar [7] has demonstrated that in a network comprising of n nodes the throughput capacity per node is O( √1n ) assuming optimal node localisation. This result includes the impact of increased interference when more nodes are introduced into the network. This leads to the conclusion that the per node capacity decreases as the number of nodes in the network increases. Various methods have been proposed to improve scalability and increase the throughput of WMNs: • the use of directional antennas to reduce the radio interference. • the use of radio power control to minimise the interference range. • the MAC protocol can minimise the number of collisions by propagation of information about use of the medium (by the use of Request-ToSend and Clear-To-Send messages). • the routing protocols can increase network throughput by the careful selection of paths

to produce minimal interference with other nodes [8]. • the use of more gateways to wired network can limit the path length and thus improve the throughput [9]. • the use of multiple frequencies provides more bandwidth and also separates parts of the network (introducing hierarchy as in backbone WMNs [10]). Routing protocols need to be aware of the consequences of the path selection process on the network. This is because the process of gathering routing information consumes bandwidth, traffic sent over selected paths also consumes bandwidth and more importantly the interference caused by the transfer of data and control frames also reduces the bandwidth. Therefore, routing protocols for WMNs need to minimise the amount of state information exchanged and also maximise the throughput of the network by an appropriate path selection process. VI

A POSSIBLE SOLUTION

The ubiquity of Wi-Fi (or 802.11) wireless devices creates a need for wide wireless coverage. Wi-Fi devices operate in unlicensed spectrum intended for industrial scientific, and medical (ISM) use. Thus, WMNs which operate in distributed manner and do not require any central administration are preferred solution. They provide multi-path routing, reliability and automatic selfdiscovery that simplifies network setup. Routing protocols for WMNs need to match the radio environment and the intended application of WMN technology. However, as has been already explained, one single solution (represented by set of routing methods selected during design process) for routing protocol does not suit all the applications. Therefore, any standards that emerge will probably leave many of the problems open, or it may only provide specific solutions for combinations of particular application/routing/radio. Currently there is considerable activity in the area of wireless municipal networks. There are many small or medium size community networks which are providing important insights about the performance of WMNs and researchers are becoming more aware of what are the major operational requirements and what are the possible solutions. Thus, if the radio range, link quality, data rates, delay and energy consumption will satisfy the users requirement, we expect that the standard for municipal WMNs will soon be defined. This will provide the incentive for the roll-out of new WMNs providing low cost Internet access for users. Other applications such as wireless mesh gaming, building automation or military communications have specific requirements which are often contradictory. Client devices of such applications may operate on variety of hardware platforms such as general purpose

computers, PDAs, mobile phones, etc. with varying capabilities. Thus we suggest that only a few routing protocols (possibly 4–5) will be required to fulfill the majority of these requirements. VII

C ONCLUSIONS

Mesh networks are resilient and can be expanded quickly. WMNs do not require extensive planning or the presence of backbone infrastructure. Individual node can quickly get up and running. One of the most important factors which determines the performance of WMNs is the routing protocol used. There is considerable industry interest in the development of routing protocols for WMNs, however no single protocol has yet emerged. Some of the protocols perform well in mobile environment but tend to have higher overhead when nodes are stationary. Other routing methods lack scalability, so they cannot provide coverage for large networks. In this paper we have presented a number of the paradigms in the design of routing protocols for WMNs. We have described the typical problems faced and the possible solutions. Moreover, we have identified that stable and scalable routing protocols can be achieved if we consider the combined effect of the underlying radio technology, the operation of routing protocol and its intended application. Only optimisation of all three elements application, routing and radio can lead to stable performance. Therefore, for the next few years the only realistic scenario is to have a variety of solutions for various applications. Maybe in the future, when the radio technology can provide sufficient bandwidth for both applications and network protocols, and when hardware of wireless devices will not impose too many restrictions on the network operation, we will finally see a single unified solution for WMNs. R EFERENCES [1] David B. Johnson, David A. Maltz, and Yih-Chun Hu. The Dynamic Source Routing protocol for mobile ad hoc networks (DSR), July 2004. Internet draft (work in progress), Available at http://www.ietf.org/ internet-drafts/draft-ietf-manet-dsr-10.txt. [2] C. Perkins, E. Belding-Royer, and S. Das. RFC 3561: Ad hoc On-Demand Distance Vector (AODV) Routing, July 2003. Available at http://www.ietf.org/ rfc/rfc3561.txt. [3] C. Santivanez and R. Ramanathan. Hazy Sighted Link State (HSLS) Routing: A Scalable Link State Algorithm. BBN technical memo BBN-TM-1301, BBN technologies, August 2001. Available at http://www.ir. bbn.com/documents/techmemos/index.html. [4] T. Clausen and P. Jacquet. RFC 3561: Optimized Link State Routing Protocol (OLSR), October 2003. Available at http://www.ietf.org/rfc/rfc3626.txt. [5] Douglas S. J. De Couto, Daniel Aguayo, John Bicket, and Robert Morris. A High-Throughput Path Metric for Multi-Hop Wireless Routing. Proceedings of the 9th

ACM International Conference on Mobile Computing and Networking (MobiCom ’03), September 2003. [6] R. Draves, J. Padhye, and B. Zill. Comparison of Routing Metrics for Static Multi-Hop Wireless Networks. ACM SIGCOMM, Portland, OR, August 2004. [7] P. Gupta and P. R. Kumar. The capacity of wireless networks. IEEE Transactions on Information Theory, 46(2), March 2000. [8] K. Jain, J. Padhye, V. Padmanabhan, and L. Qiu. Impact of Interference on Multi-hop Wireless Network Performance. ACM MobiCom, San Diego, CA, September 2003. [9] B. Liu, Z. Liu, and D. Towsley. On the capacity of hybrid wireless networks. IEEE Annual Conference on Computer Communications (INFOCOM), pages 1543– 552, 2003. [10] I. F. Akyildiz and Xudong Wang. A survey on wireless mesh networks. IEEE Communications Magazine, 43(9):23–30, September 2005.

Suggest Documents