Implications Of Quality Of Service

Published Date: 02 Nov 2017

QoS is a term widely used in the last recent years in the area of wire-based networks. QoS stands for Quality of Services and the truth is that there is much debate on what exactly QoS is supposes to mean. Most vendors implement QoS protocols having in mind specific scenarios and taking into consideration different parameters, network topologies and variables. The United Nations Consultative Committee for International Telephony and Telegraphy (CCITT) Recommendation E.800 has defined QoS as: "The collective effect of service performance which determines the degree of satisfaction of a user of the service"[63]. QOS is usually pertains to a set of service requirements that needs to be met by the network while transporting a packet stream from a source to its destination. Quality of service is not a new term in computer networks, but it did not attract much attention at during the early stages of Internet development. With the rising popularity of Quality of Service (QoS) sensitive applications such as multi-media and VoIP, the ability to provide QoS support becomes more crucial in today’s networks than it was in the past [64] .The network is expected to guarantee a set of measurable pre-specified service attributes to the users in terms of end-to-end performance such as delay, bandwidth, probability of packet loss, delay variance (jitter) etc. Power consumption and service coverage area are two other attributes that are more specific to MANets.

5.2 QoS Metrics

QoS is usually defined as a set of services that should be supported during packet transmission. A QoS enabled protocol is expected to support several metrics in terms of end-to-end throughput, delay, and jitter as well packet delivery ratio.

5.2.1 End-to-End Throughput:

End-to-End throughput, η, is defined as the ratio of the payload of effectively delivered data packets, Ped, over the elapsed time, telapsed.

η = Ped / telapsed

The basic unit of η is b/s or B/s. Effectively delivered data packets refers to data packets that are successfully delivered, excluding any duplicated packets. Since the available bandwidth in a network is fairly well known, it is helpful to obtain the actual throughput achieved which reveals the bandwidth usage efficiency. The higher the average throughput is, the better the bandwidth is utilized.

5.2.2 Delay (or Latency):

Delay, T, sometimes refers to as end-to-end delay, is the time between the originating node sending a packet and that packet reaching the destination. It may vary dramatically because of long queue time or a congested network environment.

Figure 5.1 Delay components

As shown in Figure 5.1, delay is additive in the sense that it is built up over relay nodes,

Ï„ = tS + t1 + t2 +...+ tn-1 + tn + tD

where tS and tD denote processing time at the source and destination respectively. The buffering time of a packet is of great importance for delay. If the buffering time in an individual node is set to a higher value, it could imply that packets could stay in the buffer for a long period of time when link breakages occur which will may reduce the packet dropping rate [40]. In this case, the delay is higher. On the contrary, if the buffering time is shorter, the performance of delay will improve but the packet dropping rate will increase. Delay and packet delivery ratio are traded off in different applications.

Delay can be computed in multiple layers (e.g., application layer, transport layer network layer and link layer) and thus it is layer-dependent. For the sake of synchronization, round trip delay is used in some literature while others use single trip delay.,

τ = Rt − St

where Rt and St denote time at the source and destination for a given packet respectively, assuming suitably synchronized clocks in the transmitter and receiver. In some cases, excessive delay can render some time sensitive applications such as VoIP or online gaming unusable.

5.2.3 Packet delivery ratio:

The effective delivery ratio of data packets, α, is defined as:

α = ENDP/ TNTP

where, ENDP and TNTP denote number of effectively received and total data packets respectively. Retransmission degrades the packet delivery ratio because it increases the denominator. A high packet delivery ratio is desirable, especially in MANets, since the bandwidth available is limited for wireless links.

5.3 QoS Constraints of MANETs

MANets are facing many constraints which make achievement of quality of service difficult. Compared to wired communication, MANets have several unique characteristics. To begin with, MANets rely on wireless links to transmit packets and those links are dynamic compared to wire lines since they are subject to time and location dependent signal attenuation, reflection, refraction, diffraction, and interference. Another disadvantage of wireless links is limited bandwidth.

Furthermore, the topology in MANets changes dynamically due to the unpredictable movement of nodes, which may cause network partitioning whereas in wired networks the topology seldom changes. As a consequence, protocols in MANets have to cope with movement induced path breakages. Last, but not least, some devices in MANets are battery powered and thus energy consumption must also be taken into consideration in the network design. In wired networks, devices always have enough power and the energy constraint is rarely considered.

Due to the significant difference in MANets, the mechanisms for wired networks cannot be mapped to MANets directly. QoS provision in MANets is quite challenging and it involves actions in different layers within which the network layer plays a crucial role. The routing protocol in the network layer not only has to find a path, if any, that can satisfy QoS requirements at the beginning of a session but also needs to react to mobility induced route breakages. Numerous efforts have been devoted to addressing this problem, leading to the introduction of a series of QoS provision protocols with QoS metrics support.

However, it is observed that many applications in real world usually have more than two QoS constraints simultaneously that are, sometimes, contrary [65]. To design a single protocol with two or more QoS constraints is known to be a NP-complete problem and the time to solve a NP-complete problem using algorithms available currently increases dramatically as the size of the problem increases.

Routing protocols with diverse QoS metric support are application-dependent which means a new algorithm has to be implemented as the application or environment changes. Last, but not least, the relative importance of QoS metrics in applications is neglected in much literature.

Meanwhile, in all there are three new constraints which MANet has to face as compared to conventional wired network. These constrains are: the Bandwidth Constrains, since a MANet has usually poor bandwidth resources, the Dynamic Topology of the MANet, since nodes are continually changing location, connecting and disconnecting from the network making connections many times unreliable, and c)the Limited processing and Storing capabilities of mobile nodes.

The issues can be further elaborated as follows:

5.3.1 Unpredictable Link Properties:

Wireless media is very unpredictable. Packet collision is intrinsic to wireless network. Signal propagation faces difficulties such as signal fading, interference, and multi-oath cancellation. All these properties make the measures, such as bandwidth and delay of a wireless link unpredictable

5.3.2 Node Mobility:

Mobility of the node creates dynamic network topology. Links will be dynamically formed when two nodes come into the transmission range so each other and are torn down when they move out of range.

5.3.3 Limited Battery Life:

Mobile devices generally depend on finite battery sources. The resource allocation for QoS provisioning consider the residual battery power and the rate of battery consumption corresponding to the resource utilization. Thus all the techniques for QoS provisioning should be power-aware and power-efficient

5.3.4 Hidden and Exposed Terminal Problems:

The hidden terminal problem happens when signal of two nodes, say A and B, which are out of the transmission range of each other collide at a common receiver say C. With the same nodal configuration, an exposed terminal problem will result from a scenario where node B attempts to transmit data (to someone other than A&C) while node C has been transmitting to node A. In such a case, node B is exposed to the transmission range of node C and thus defers its transmission even though it would not interfere with the reception at node A.

5.3.5 Route Maintenance:

The dynamic nature of the network and the changing behavior of the communication medium make the precise maintenance of network state information very difficult.

5.3.6 Security:

Security can be considered as a QoS attribute. Without adequate security, unauthorized access and usages may violate the QoS negotiations. The nature of broadcasts in wireless networks potentially results in more security exposures. The physical medium of communication is inherently insecure.

5.4 QoS architecture

In Figure 5.1, we show our proposed QoS architecture, which includes all networking layers from the application layer to the MAC layer. The bold lines indicate the flow of data packets and the narrow lines indicate the flow of control packets. Each layer’s features are described in detail as follows:

5.4.1 Application Layer:

Applications can be categorized into real-time and non-real-time applications based on their sensitivity to packet delay. Real-time applications have strict requirements on the packet delay. Therefore, packet retransmission is not allowed. The applications that fit into this category are on-line live movies and video conferencing. Many video compression technologies, such as MPEG-4, H.263, and multiple-description coding, can compress video with different coding rates to meet different channel conditions. In addition, most of these compression schemes have error resilience features to recover the video frame, if some packets are lost. Thus, choosing the right coding rate to compress the video is important, and some reasonable packet loss is acceptable. For this purpose, we need an efficient model which highly improves the packet delivery ratio.

In this thesis, an attempt is made to achieve high efficiency using "Bandwidth Estimation" technique. On the other hand, for non-real-time applications such as Email and FTP, packet delay is not a big issue, and packet delivery is guaranteed by explicit acknowledgements in the transport layer.

Fig 5.2 : QoS Architecture[79]

5.4.2 Transport layer:

UDP and TCP are two transport layer protocols widely used in wired networks. UDP has no congestion control scheme to react to network congestion. Applications that use UDP as the underlying transport protocol to transmit packets can easily overwhelm the network with data, which results in a considerable amount of wasted energy and bandwidth in transmitting packets that will be dropped due to congestion. Therefore, some pre-dropping of UDP packets should be investigated to react to congestion. TCP has an inherent congestion control scheme, so congestion control is not a problem. However, TCP’s performance should be optimized to adjust the TCP window, which requires feedback information from the lower network layers. Therefore, some information from the packet queue and the routing layer should be sent to the transport layer for performance optimization.

5.4.3 Network layer:

To support QoS, the routing protocol should have an embedded scheme such as call admission or adaptive feedback that is designed to support QoS. At the same time, non-QoS-aware routing that is targeted at finding a feasible path should be offered as well. For QoS-aware routing, information about the current network status is provided to the application for performance optimization. Also, the routing layer should get enough channel information from the lower layers so that the admission/adaptive scheme can be performed based on the network status. Therefore, two cross-layer designs should be implemented in QoS-aware routing. One is to obtain the network resource information from lower layers, and the other is to send the network status to the applications. To offer QoS to the applications, resource reservation should be incorporated. An RSVP-type signaling scheme is not desirable in MANets due to its high overhead. Therefore, in-band and soft resource reservation (i.e., best effort rather than guaranteed reservations) should be done during the route discovery phase and during route maintenance. The transmissions that occur between the breakdown of old routes and the set up of new routes will severely affect the QoS provided by the network. Therefore, some prediction of route breaks should be incorporated.

Overall, QoS-aware routing should have the following features that traditional routing does not support:

Obtain resource information from lower layers;

Offer bandwidth information to applications;

Incorporate resource reservation schemes;

Predict route breaks.

5.5 Link Layer

The link layer needs to discriminate the different priority packets and schedule packet delivery according to priority levels. The service differentiation should be completed in the packet queue through queue management and in the MAC layer through a MAC discriminator and priority classifier.

5.5.1 Queue Management:

The aim of queue management is to schedule the different priority packets. Real-time data should have higher priority to be sent to the channel compared with packets such as FTP and Email. Therefore, real-time data will be put in front of the non-real- time data in the packet queue. When the network is congested, the last packet in the packet queue will be dropped. Therefore, incorporating queue management will reduce the possibility that real-time packets are dropped in the packet queue when the network is congested. Thus, packet delivery ratio can be improved. Also, the packets whose delay has already exceeded the applications requirement should be eliminated from the packet queue before transmission to save the transmission of packets that will be useless to the receiver. If different flows go through the same host, it is easier to do the priority regulation in the packet queue than in the MAC layer.

5.5.2 MAC Discriminator:

The main function of the MAC discriminator is to differentiate data packets and control packets that arrive from the wireless channel. Data packets are sent to the network layer; ARP (address resolution protocol) packets go to the queue directly; MAC packets, such as the RTS, CTS, and ACK packets used in IEEE 802.11, stay in the MAC layer; and the bandwidth estimation control packets are sent to the bandwidth estimation module for use in the routing layer’s admission/adaptive scheme.

5.5.3 Priority Classifier and Packet Scheduler:

To offer service differentiation in a distributed Ad hoc network, real-time packets should be granted higher priority to capture the channel. The priority classifier differentiates the different data packets that arrive from the packet queue and directs the packet scheduler to schedule the packet delivery based on the priority level of the current packet.

5.6 Bandwidth Estimation

Conventionally, network bandwidth corresponds to width of spectrum of electromagnetic waves employed in the network. In some cases, it is used to describe propagation characteristics of communication channel. However, for Ad hoc networking scenario, it the maximum rate of data transfer of data link established temporarily among communicating nodes in the network [66].

As already mentioned, available bandwidth of a network is an important parameter of quality of service. Nowadays, several applications generate multimedia data flows that need networks with sufficient and reliable bandwidth availability. These applications may benefit from a quality of service (QoS) support in the network in terms of available bandwidth. That is why this domain has been extensively studied and more and more QoS solutions are proposed for Ad hoc networks. However, the term QoS is vague and gathers several concepts. Some protocols intend to offer strong guarantees to the applications on the transmission characteristics, for instance bandwidth, delay, packet loss, or network load. Other solutions, which seem more suited to a mobile environment, only select the best route among all possible choices regarding the same criteria. In both cases, an accurate evaluation of the capabilities of the routes is necessary. Most of the current QoS proposals leave this problem aside, relying on the assumption that the link layer protocols are able to perform such an evaluation. However, these protocols do not have such provision. Hence, the network resource evaluation problem offers substantial research challenge as not only it needs to take into account several phenomena related to the Ad hoc networking scenario & wireless environment but also dependent on less measurable parameters such as the node mobility.

Hence, throughout this research, the focus on one of the fundamental resources: bandwidth. Estimating the remaining bandwidth at a given time and in a given part of the network is tricky because, in a wireless network, the medium is shared between close nodes. Consequently, computing the available bandwidth between two neighbor nodes necessitates an accurate identification of all potential contenders and their impact on each other not only on at the emitter’s side and all potential scramblers as they compete and struggle for possession of bandwidth at receiver’s side. Information about nodes’ utilization of the shared resource should, therefore, be gathered and composed to derive the amount of free resources. Both tasks are usually difficult to realize.

The task of bandwidth estimation becomes difficult in the networks with sparsely distributed nodes. In such networks, two nodes may share the medium without being able to directly exchange information. Moreover, data packet being transferred in such network may need to hop among multiple intermediate nodes before reaching to their destinations. The task become even more difficult as each host has imprecise knowledge of the network status and links change dynamically. Therefore, an effective bandwidth estimation scheme is highly desirable.

Bandwidth estimation can be done using various methods; the one of the way of bandwidth estimation is a cross-layer design of the routing and MAC layers, and in second the available bandwidth is estimated in the MAC layer and is sent to the routing layer for admission control. Therefore, bandwidth estimation can be performed in several different network layers, as shown in Figure 5.2. In a distributed Ad hoc network, a host’s available bandwidth is not only decided by the raw channel bandwidth, but also by its neighbor’s bandwidth usage and interference caused by other sources, each of which reduces a host’s available bandwidth for transmitting data. Therefore, applications cannot properly optimize their coding rate without knowledge of the status of the entire network. Thus, bandwidth estimation is a fundamental function that is needed to provide QoS in MANets.

However, there is no standardized method or mechanism for accurate evaluation of extent of resources currently available at any channel under consideration. However such type of evaluation will be important for network applications where bandwidth is limited. However for multihop Ad hoc networks, this type of evaluation is not easy to obtain. Hence in spite of many research efforts, much work need to be done for estimation of available bandwidth and continues to be an important issue in Ad hoc networking research field.

Research work pertaining to estimation of available bandwidth has generated large amount of literature both in Ad hoc network research community. There are so many ways of classification of bandwidth estimation techniques. However, these techniques can be broadly divided into two categories i.e. active and passive approaches. Active approaches: In these methodologies, an available bandwidth is estimated by observing emission of dedicated end-to-end probe packets.

5.6.1 Active Bandwidth Estimation Techniques:

In this technique an available bandwidth is estimated by deliberately introducing test data flow of dedicated end-to-end data packets. The technique is called active bandwidth estimation due to active involvement of estimator in the process by means of its own data packets rather than passive observation normal data packet flow.

The probe packet size and extent of cross-traffic have great influence on the estimated bandwidth in such networks than in wired networks. Hence, these techniques are also very sensitive to the evaluation parameters and many times lead to unacceptable results in a wireless scenario. The active techniques mentioned here have two major shortcomings with respect to multi-hop Ad hoc networks.

Firstly, volume of probe packets becomes significant in comparison with normal traffic when bandwidth the estimation is done on many nodes simultaneously. The probe packet traffic may interfere with data traffic and adversely modify the estimation. Secondly, an end-to-end evaluation technique may not be as effective as a local technique in high mobility scenario. Contrarily, local detection of bandwidth may be more effective for efficient updating of routing information especially in the situations of frequent alteration of extent of available resources or high mobility of nodes.

Many of these methods measure bandwidth available for communicating nodes by transmitting packets of identical size from a source to destination. The source gradually increases the probe packet emission rate. Measurements of the parameters of this data flow are carried out at the receiver’s side and then transformed into an estimation of the end-to-end available bandwidth. Many of the protocols like as SLoPS [67] or TOPP [68] are of this type. These two approaches mainly differ in the manner in which they alter probing packet delivery rate and nature of parameters measured on the probing packet flow. However, the traffic of these data packets may increase to One of drawback of such techniques is the fact that, flow due to traffic of probing packets may interfere normal data packet flow and hence adversely affect correct evaluation of available bandwidth.

Li et al. [69] presents mechanism of detection of presence of network congestion by observing probe packet delay. Increase in delay beyond theoretical maximum delay indicates that the congestion has occurred. They propose a mechanism measurement of medium utilization and estimation of channel capacity from this channel usage ratio. Based on the TOPP method, the authors of DietTOPP [70] measured the accuracy of such methodologies in wireless networks.

5.6.2 Passive Bandwidth Estimation Techniques:

As direct contrast to the active techniques, these techniques employ assessment of only local information of bandwidth utilization. In one of the approaches, channel usage is monitored by sensing and monitoring physical channel. These mechanisms exchange information by one-hop broadcasts as such information can be send along with Hello messages. This information is used by many routing protocols determine local topology.

A dynamic bandwidth management scheme for single-hop Ad hoc networks is proposed in [71]. Here, one node in the network hosts the Bandwidth Manager process, which is responsible for evaluating the available bandwidth in the cell and for allocating the bandwidth to each peer. Each node may ask the Bandwidth Manager for an exclusive access to the channel during a proportion of time using dedicated control messages. As the topology is reduced to a single cell, the available proportion time-share is computed by this entity considering that the total load is the sum of the individual loads. The available fraction of time may then be translated into an available bandwidth by considering the capacity of the wireless link, called total bandwidth in this paper, which is deduced from a measurement of the data packets’ throughput. This approach can be considered as passive as very few control packets are exchanged, usually of small size. However, this solution is adapted to network topologies where all the nodes are within communication range but cannot be directly used in multihop ad hoc networks.

Even if the election, the synchronization, and the maintenance of several Bandwidth Managers may represent a significant cost in large distributed networks, similar measurements may be employed. When a node desires to estimate the bandwidth available in its vicinity, the intuitive approach consists in monitoring the channel over a given time period and to deduce from this observation the utilization ratio of the shared resource. The method proposed in [72] uses such technique and adds a smoothing factor to hide transient effects. The QoS routing protocol designed in this paper is based on a simple estimation of the available bandwidth by each node and does not consider any interfering nodes.

QoS-AODV [73] also performs such a per-node ABE. The evaluation mechanism constantly updates a value called Bandwidth Efficiency Ratio (BWER), which is the ratio between the numbers of transmitted and received packets. The available bandwidth is simply obtained by multiplying the BWER value by the channel capacity. This ratio is broadcasted among the one-hop neighbors of each node through Hello messages. The bandwidth available to a node is then inferred from these values as the minimum of the available bandwidths over a closed single-hop neighbourhood. QoS-AODV, therefore, considers not only the possibility to send a given amount of data but also the effect of the emissions of a node on its neighbourhood.

Bandwidth reservation protocol called Bandwidth Reservation under InTerferences influence (BRuIT) [74]. This protocol’s ABE mechanism takes into account the fact that, with the IEEE 802.11 standard, the carrier sense radius is larger than the transmission range. In other words, emitters share the bandwidth with other nodes they cannot communicate with. Studies have shown that this carrier sense radius is at least twice the communication radius. To address this issue, each node regularly broadcasts to all its immediate neighbours’ information about the total bandwidth it uses to route and emit flows (deduced from applications and routing information) and its estimated available bandwidth. It also transmits similar information concerning all its one-hop neighbours, propagating such information at a two-hop distance. Each node then performs admission control based on this two-hop neighbourhood knowledge. When the carrier sense radius is equal to twice the communication radius, the authors have shown that two-hop communication represents the best compromise between estimation accuracy and cost [75].

Making the same observation, Yaling and Kravets [76] proposed the Contention Aware Admission Control Protocol (CACP). In this framework, each node first computes its local proportion of idle channel time by monitoring the radio medium. Then, the authors propose three different techniques to propagate this information to the greatest number of nodes within the carrier sense area. First, similarly to BRuIT, they propose to include the information in Hello messages to reach the two-hop neighbourhood. Second, they propose to increase the nodes’ transmission power; however, this emission power is often limited by regulations and this technique may therefore only be applicable when power control is used for regular transmissions. Finally, receiving nodes can also reduce their sensitivity in order to decode information coming from farther away, which depends on the quality of electronics and on the signal modulation. Similarly to [77], the authors also point out the existence of intraflow contention. When a flow takes a multihop route, successive routers contend for channel access for frames belonging to the same flow. It is thus important to take into account at least the route length when performing admission control. Ideally, the exact interactions between nodes along a path should be identified and considered.

Finally, the AAC protocol, proposed in [78], makes each node consider the set of potential contenders as a single node. It measures the activity period durations and considers that any such period can be seen as a frame emission of the corresponding length. With this mechanism, collisions and distant emissions are also considered when computing the medium occupancy. Based on this measurement, each node is able to evaluate its available bandwidth. It exchanges this information with its neighbors to compute the bandwidth on each link, a link being defined as a pair of nodes. This value is defined as the minimum between the available bandwidths of both ends. AAC also takes into account the intraflow contention problem mentioned above.

In this research, I present an enhanced mechanism for estimation of available bandwidth in IEEE 802.11-based Ad hoc networks. The mechanism is discussed in detail in chapter on methodology i.e chapter no 7.