QoS Architecture For A Mobile Ad Hoc Network.pdf

QOS ARCHITECTURE FOR A MOBILE AD HOC NETWORK Richard C. Bernhardt J. Bibb Cain William A. Windham Harris Corporation Government Communications Systems Division Melbourne, FL ABSTRACT Dynamic allocation of network resources in packet-based mobile ad hoc networks requires the operation of autonomous mechanisms that can provide assurance that the user or host application can reasonably expect a satisfactory quality-of-service (QoS). In general, a combination of mechanisms is required to operate concurrently and in real-time to meet this objective. This paper discusses an architectural framework for a mobile ad hoc node, where the functionality required to achieve QoS is achieved by a combination of functional layers and cross-layer control methods. The framework enables functionalities to support QoS at the application, transport, network, MAC, and physical levels by adaptation to application traffic loads, network conditions, topology, and radio link performance. The QoS framework described provides a basis for many possible implementations. Examples of its application that illustrate key functionalities are discussed. These include: adaptation of physical layer attributes to enhance network fabric capacity, source to destination path selection to enhance end-to-end performance, and admission control to facilitate operation. INTRODUCTION MANET technology is seeing a wider range of requirements and applications in terms of size, operational environment (mobility and propagation environment) and offered traffic characteristics. Multiple QoS mechanisms at different layers need to be coordinated. Additionally, these mechanisms may be applied in different combinations. A framework is needed to provide a unified solution that can use one or more of these mechanisms in a given implementation. This paper discusses an architectural framework to achieve QoS by the use of cross-layer mechanisms that cooperate but also maintain a high degree of modularity. The framework enables functionalities to support QoS at the application, transport, network, MAC, and physical levels 1-4244-1513-06/07/$25.00 ©2007 IEEE

by adaptation to changes in application data traffic loads, network conditions, topology, and radio link performance. The use of multiple functional and layers and cross-layer control is discussed, and several key application examples are presented at the physical, network, and applications levels for the patents on which this paper is based [1][2][3]. There has been previous work done on implementing QoS in MANETs [4][5][6]. Some of the previous work has addressed the problem at the network level by utilizing or adapting conventional QoS techniques such as DiffServ [7]. PROBLEM DESCRIPTION Figure 1 shows a scenario of a tactical MANET consisting of a group of warfighters requiring reliable communications with a minimum degree of predictability in a particular mission scenario. The challenge is to achieve an integrated solution to obtaining and maintaining QoS for their network under a range of expected movement within a tactical environment. Current routing algorithms (e.g., OLSR[8], AODV[9]) provide only one route, usually with minimum number of hops; it would be advantageous if each node could more performance qualified routes, which would enhance the communications availability and reliability needed to successfully support the mission and safeguard the members of the MANET. Nodes connected by a single link are at risk of being cutoff due to variable signal and interference from changing traffic conditions as well as movement. Current solutions do not provide systematic adaptation of physical level radio characteristics such as power control, power saving, required coding strength, maximum required data rates, and antenna gain and pattern control. The rest of the paper discusses the functionality of the proposed QoS framework, performance aspects, measures associated with topologies and end-to end networks, and several example applications of the framework.

1 of 6

APPLICATION TOS

ADMISSION CONTROL

SESSION / TRANSPORT QoS SUPPORT (INCL. REL. MULTICAST) QoS PACKET CODING PATH SELECTION, & CONNECTIVITY MANAGEMENT

Figure 1. Scenario

MANET QoS FRAMEWORK

PATH LIST

METRIC PROC.

QoS FORWARDING (UNICAST / MULTICAST)

Figure 2 is a layer view of the QoS framework. It operates as a multi-layer protocol hierarchy. Admission control is performed at the application layer, with the establishment of type of service (TOS) according to the particular application’s type and priority. For example, time sensitive data, such as video or audio data, requires a higher priority than best-effort data. The session and transport layer provides for reliable delivery if required, for unicast or multicast. The QoS packet-coding layer can be used to increase reliability, if additional overhead can be tolerated. The QoS forwarding layer may select between unicast and multicast communications modes. The QoS Traffic and Queue Management layer performs traffic shaping, policing, and network queue management.

QoS TRAFFIC & QUEUE MANAGEMENT

TOPOLOGY AND NEIGHBOR DISCOVERY, MANAGEMENT, RAL CONTROL, AND QoS METRIC DATA COLLECTION

RADIO ADAPTATION LAYER MAC

MAC

MAC

PHY

PHY

PHY

Figure 2. MANET QoS Framework

Figure 3 is a representation of concurrent operation of three key processes in the operation of the QoS framework that are further addressed in this paper. QoS PROCESSES

QoS METRIC COMPONENTS

ADMISSION CONTROL

At the path selection layer, end-to-end paths to a given destination are available based on a QoS metric. At the QoS traffic and queue management, traffic shaping, queuing discipline, packet prioritization, and flow control are performed. The radio adaptation layer (RAL) provides interfaces to radio devices (MAC/PHY) that transmit and receive data over a link that supports the selected path, which may use multiple hops. This layer performs radiospecific control functions for power control, modulation type and rate, and antenna configuration. The RAL also collects link-specific information for QoS metric establishment.

APPLICATION TOS

SERVICE SENSITIVITY APPLICATION TYPE APPLICATION PRIORITY

SESSION / TRANSPORT QoS SUPPORT (INCL. REL. MULTICAST) QoS PACKET CODING PATH SELECTION, & CONNECTIVITY MANAGEMENT

PATH OPTIMIZATION

NO. OF PATHS NO. OF HOPS PER PATH PATH DELAY

QOS FORWARDING (UNICAST / MULTICAST) QOS TRAFFIC & QUEUE MANAGEMENT

Implementation of the QoS metric, which is also used for control purposes, is based on data collected by a node from its radio environment. Some of that data is processed for the purpose of end-to-end path selection and management. Other data collected for the metric comes from the applications that are served.

FABRIC CAPACITY MAXIMIZATION

RADIO ADAPTATION LAYER MAC

MAC

MAC

PHY

PHY

PHY

NO. OF NEIGHBORS NEIGHBOR STATE PACKET ERRORS BIT ERROR RATE JITTER AVAILABLE BANDWIDTH AVAILABLE TX POWER

Figure 3. Concurrent QoS Processes

The switch icons on the left hand side of the diagram in Figure 3 indicate that the functionalities of those layers may be controlled by an established policy, dynamic network management not directly part of the QoS 2 of 6

framework, or by other implementations of the framework not described in this paper.

Maximizing fabric capacity is a dynamic ongoing process, and improves end-to-end performance [10].

PERFORMANCE ASPECTS

FRAMEWORK APPLICATIONS

Figure 4 shows a model of a 7-node MANET where it is assumed that each node has three adjacent nodes that can be reached directly, i.e., without multi-hopping. The exception is node 4, which is one hop away from all the others.

Using physical layer adaptation to enhance the throughput of the local topology or neighborhood can enhance network capacity. In the flowchart in Figure 6, each node adjusts its transmitted data rate to increase the potential maximum network capacity, while achieving a “sufficient” number of neighbors with reliable links, and minimizing co-channel interference. It is assumed that transmitter power is held constant, and path loss is homogeneous, isotropic, and varies as dn. Note that signal bandwidth will increase as the data rate increases.

Figure 4. 7-Node MANET Model

If these nodes were part of a larger MANET as shown in Figure 5, they would constitute a neighbor group that would support through traffic with some maximum aggregate throughput (bps) subject to a given transit delay. The maximum aggregate throughput of the group, with its corresponding delay, provides a measure of network fabric capacity. This is the case for a regular grid, but also for a network where nodes are randomly located with a high density. AGGREGATE OFFERED TRAFFIC TO NEIGHBOR GROUP

AGGREGATE NEIGHBOR GROUP THROUGHPUT

TRANSIT DELAY

Figure 5. 7-Node Neighborhood within a Large MANET

Figure 6. Adaptation of Physical Layer to Enhance QoS

An increase in the maximum achievable network fabric capacity is made possible by physical level adaptation in the form of increasing transmitter power, assuming interference is sufficiently mitigated by the radio MAC. This paper does not address a specific transmission format, media access method, or channelization scheme. However, 3 of 6

when multiple channels are available in operation, adaptation via channel selection is an important physical layer adaptation to mitigate interference. [11][12] Consider the two source to destination paths shown in Figure 7: the one-hop path from S to D, and the two-hop path from S to I to D. Using standard link budget and path loss model techniques, we assume that the maximum possible data rate that can be sent (R) is proportional to the maximum transmitter power (P) and inversely proportional to the nth power of the hop distance, d, which is a common model of average path loss in a terrestrial propagation environment. In this case,

R

P dn

n

1 C1 R1 / 2 R1 / 2    22 n/2 C2 R2 R1 / 2

(3)

where d2 = d12 for a square grid. Following the same reasoning for a hexagonal grid (see Figure 8), the expression for the relative maximum possible data rate improvement and corresponding improvement in maximum relative capacity is: n

1 C1 R1 / 3 R1 / 3    32 n/2 C2 R2 R1 / 3

(1)

(4)

D d2

where n is the path loss exponent. Here we assume that the modulation (e.g., QPSK) is kept constant, and that the bandwidth is changed as needed. In the case of free space path loss modeling, n is 2. (The path loss exponent is generally greater than 2 for VHF or UHF frequencies).

S

d1

d1

In the next two examples, we assume that the transmitter power is constant but the data rate can be adaptively adjusted to a rate that can be supported reliably over the link of a specified distance. For the square grid in Figure 6, the ratio of the maximum data rate that can be sent over distance d2 as compared to distance d1 is: Figure 8. Hexagonal Grid Neighborhood

R1 2 n/2

(2)

Figure 9 is a graph of C1/C2 for square and hexagonal grid neighborhoods as a function of the path loss exponent, n. IMPROVEMENT IN NETWORK FABRIC CAPACITY (C1/C2)

R2 

D d2

S

d1 d1

10

1

Figure 7. Square Grid Neighborhood

2

Since the two-hop path uses twice the channel access resources (think of needing to access the channel twice using a TDMA channel access protocol) as the one-hop path, it is equivalent to saying that the ratio of the improvement of the corresponding maximum potential network capacities is: 4 of 6

3

4

5

PATH LOSS EXPONENT (n) Square Grid

Hexagonal Grid

Figure 9. Improvement in Relative Maximum Network Fabric Capacity from Data Rate Adaptation

6

This example assumes that power is fixed and increasing signal bandwidth while keeping modulation constant can increase the data rate. The dual approach keeps the data rate on each link constant, while adaptively controlling transmitter power on each link. Working through this example in the same fashion produces a similar result if the objective is to minimize network transmitted power to achieve a given network fabric capacity. Figure 10 is a flow model corresponding to the MANET shown in Figure 4. It is valid during the given epoch in which the topology established by the mutual coverage areas of the nodes does not substantially change.

Data for QoS Metric

As an example, if a new traffic flow is applied by node 6, with node 3 specified as the destination, its TOS is checked and an admission control decision is made. If the flow can be admitted, node 6 identifies available paths to node 3. Node 6 examines the priority and type of the new application flow, and determines which paths satisfy QoS criteria. A path is then selected, which in this example is the 4-hop path 6 to 1 to 2 to 4 to 3 shown in Figure 12. Note that even though that path may not be the shortest in terms of number of hops, it does meet the admission control, path, and neighborhood link connectivity criteria.

PATH1

PATH2

SRC PATH7

6

DEST 3

PATH14

PATH15

Figure 11 - Multi-hop Route from Source to Destination Figure 10 - Flow Model Conditioned by Admission Control

There are 15 distinct acyclic multi-hop paths from source (node 6) to destination (node 3), which are listed in Table 1. Associated with each path is a QoS metric, and node 6 also has metrics related to the link qualities of neighbor nodes 1, 4, and 7. Table 1. Source to Destination Paths

PATH 643 6123 6753 6143 6743 6423 6453 61243 67543 64123 64753 614753 674123 6124753 6754123

No. of Hops 2 3 3 3 3 3 3 4 4 4 4 5 5 6 6

Over a period of time, new flows occur, and existing flows may terminate. During that time, each node’s concurrent QoS processes update QoS metrics accordingly, so that future path selection can proceed while all nodes maintain their overall QoS. Figure 12 is an illustration of performance regions of a MANET in terms of the relationship between fabric capacity and link transmission quality, and how the QoS framework could operate in support of maintaining QoS at the lower layers, which in turn enhances the sustaining of good end-to-end performance. First, note that TOS-based admission control admits only those traffic flows that can be accommodated by the MANET, given its basic resources. In Region 1 (R1), performance is typical of a wired or conventional network, and conventional QoS mechanisms are sufficient. There is an outer band of R1 where performance may be acceptable for a MANET, and conventional QoS mechanisms may still be able to handle performance management in low mobility and benevolent propagation environments.

5 of 6

In the opposing region R4, operating conditions are so poor, that without a purposeful change in topology, frequency channel, or access to relay nodes, the MANET is not capable of sustaining communications. In R2 and R3, the mechanisms of a design based on the framework discussed in this paper, act to maintain or improve operation according to local conditions. Then, physical layer adaptation, such as that discussed in the prior section, attempts to improve transmission quality and acquire enough neighbors, to restore performance according to concurrent actions based on the QoS metrics defined in a detailed design. Two performance adjustment trajectories are shown in Figure 12. In the first trajectory example, nodes in a given neighborhood experience are experiencing poor transmission quality (BER), which is detected by excessive packet decoding errors. The physical layer QoS mechanisms respond by either lowering the modulation level to reduce noise bandwidth, increase the amount of coding (incurring some addition overhead), and/or modifying the antenna pattern, if possible, to increase gain in the desired directions.

R2

R1

NEED IMPROVED LINK QUALITY

“WIRED” NETWORK PERFORMANCE AND QoS

1

Decrease modulation level, change antenna pattern, strengthen coding

R4 POOR PERFORMANCE (OPPORTUNISTIC COMMUNICATIONS ONLY)

PE AC R CE FO P RM TAB A LE N CE

Increase TX power for increased coverage, increase data rate, or modulation level to increase capacity

R3

2

MANET FABRIC CAPACITY (BPS)

In the second trajectory, local nodes attempt to increase fabric capacity by increasing transmission power to acquire more neighbor nodes, increase data rate, preferably by increasing modulation level (to keep the noise bandwidth constant), or use any available relays to increase the number of possible “through routes”.

NEED MORE NEIGHBOR NODES OR RELAYS

0



TRANSMISSION QUALITY (BER)

0

Figure 12. Fabric Capacity vs. Transmission Quality

CONCLUSION This paper discussed an architectural framework for a mobile ad hoc node, which has the functionalities required to achieve sustainable QoS for a MANET. This is done with a combination of functional layers, concurrent operation of integrated cross-layer control methods, and the use and sharing of QoS metrics at multiple layers in the protocol stack of the node. Several examples of its possible use were discussed, with many more being possible. REFERENCES [1] J. Cain, R. Bernhardt and W. Windham, “Mobile Ad Hoc Network (MANET) providing Interference Reduction Features and Related Methods”, Patent Number 7,068,605, 2003. [2] J. Cain, R. Bernhardt and W. Windham, “Mobile Ad Hoc Network (MANET) with Quality-Of-Service (QoS) Protocol Hierarchy and Related Methods”, Patent Number 7,079,552, 2003. [3] J. Cain, R. Bernhardt and W. Windham, “Mobile Ad Hoc Network (MANET) providing Connectivity Enhancement Features and Related Methods”, Patent Number 7,085,290, 2003. [4] M. Brahma, K. W. Kim, A. Abouaissa and P. Lorenz, “A New Approach for Supporting QoS in Mac Layer over MANETs”, Proceedings of the 2005 Systems Communications (ICW’05). [5] Xiaohua Jia, Deying Li and Dingzhu Du, “QoS Topology Control in Ad Hoc Wireless Networks”, Proceedings of the 2004 IEEE INFOCOM. [6] Kui Wu and Janelle Harms, “QoS Support in Mobile Ad Hoc Networks”, Crossing Boundaries – an interdisciplinary journal, Vol. 1, No 1, Fall 2001. [7] S. Blake, “An Architecture for Differentiated Services”, IETF RFC2475, December 1998. [8] T. Clausen, P. Jacquet, “Optimized Link State Routing Protocol (OLSR)”, IETF RFC3626, October 2003. [9] C. Perkins, E. Belding-Royer, July 2003, “Ad hoc OnDemand Distance Vector (AODV) Routing”, IETF RFC3561. [10] Piyush Gupta and P. R. Kumar, “The Capacity of Wireless Networks”, IEEE Transactions on Information Theory, Vol. 46, No. 2, March 2000. [11] R. C. Bernhardt, "User Access in Portable Radio Systems in a Co-Channel Interference Environment", IEEE Journal on Selected Areas in Communications, Vol. SAC-7, No. 1, January 1989. [12] R. C. Bernhardt, "Time-Slot Management in Frequency Reuse Digital Portable Radio Systems", IEEE Vehicular Technology Conference, Orlando, Florida, May 1990 and IEEE Transactions on Vehicular Technology, Vol. 40, February 1991, pp. 261-272.

6 of 6