AN ADAPTIVE GATEWAY DISCOVERY IN HYBRID MANETS F. D. Trujillo, A. - - PowerPoint PPT Presentation

AN ADAPTIVE GATEWAY DISCOVERY IN HYBRID MANETS

• F. D. Trujillo, A. Triviño, E. Casilari and A. Díaz-Estrella

Department of Electronic Technology University of Malaga

• A. J. Yuste

Department of Telecommunication Engineering University of Jaen

Contents

1.- Introduction 2.- Previous studies 3.- Adaptive Gateway Algorithm (AGW) 4.- Performance evaluation 5.- Conclusions

– Necessity of an access router and an internet gateway Modified Router Advertisement messages (MRA) – Interconnection of Mobile Ad hoc NETwork (MANET) and Internet to increase the network capacity

Introduction (I)

– The emission of MRA messages can be achieved by three diferents schemes:

• Proactive mechanism: the internet gateway disseminates

the message periodically

• Reactive mechanism: these messages are generated only
• n demand as reply of a MRS message
• Hybrid mechanism: combines both previous schemes

(MRA messages to devices nearby to the internet gateway)

Introduction (II)

– We will focus on the proactive gateway method – Procedure of performance: 1. A mobile node receives the MRA message 2. This mobile node updates its route entry 3. It rebroadcasts the MRA message – The interval of generation of MRA messages (T period) affects the network:

• Low value: the limited MANET resources can be cousumed
• High value: former routing information can be stored in the nodes

– It is necessary to choose very carefully this advertisement period

– In proactive mechanism, better performance (packet delay) is obtained (but with a high bandwidth use) – The goal: decrease control traffic – Algorithm based on the estimation of network conectivity from the percentage of nodes located in the transmission range of the internet gateway:

• High number of neighbours: shorther routes are required for external

communications (longer lifetimes) T higher

• The gateway analyzes the number of MRA messages and changes T to avoid

emission of excessive routing packets.

• This tuning process is supported by a control system configured by means of

statistical properties (after explained)

Introduction (and III)

Previous studies (I)

– The interval of emission of MRA messages (T) must be adapted to the network conditions – Related algorithm:

• Maximal Source Coverage (MSC): T is a fixed value and the internet gateway

sents the next MRA message with TTL = minimun number of hops

• Regulated Mobility Degree (RMD): the MRA messages broadcasting depends
• n the number of active sources and the number of intermediate nodes
• Use of an auto-regressive filter to adjust both T and TTL, simultaneously.

Necessity of monitor de traffic load in internet gateways

• Dynamically tuning of T by means of the estimation of reactive route

solicitations from the nodes by means of auto-regressive filter

• Adaptive Gateway Algorithm (AGW): the adaptation of T is

based on the number of MRA messages retransmitted by the gateway’s neighbours

– Until now: the T period is fixed to a constant value – However: the optimum value depends on the network conditions (load, number

• f traffic sources, the node mobility, etc)

– From now: the T period is adjusted with regard to the number of received MRA messages which are retransmitted by the gateway’s neighbours:

• Many MRA messages received: all these nodes have updated the routing entry

to the internet gateway the T could be incremented

• Few MRA messages received: the T must be decreased to guarantee that nodes

keep valid route to the internet gateway

Adaptive Gateway Algorithm (I)

Adaptive Gateway Algorithm (II)

Output function of control system to adapt T period – Input: the number of MRA messages received by the internet gateway – Output: the T period – The measurement of received MRA messages is carried out every period of T – Some necessary simulations to justify the selection of a linear function like measurement of the network connectivity

5 10 15 20 25 30 35 40 4 6 8 10 12 14 16 18 20 22 24 MRA received Real node mobile close GW

Adaptive Gateway Algorithm (III)

MRA messages received versus node mobile close to the gateway – The probability p that a node is near a gateway can be calculated and it depends

• n the node number, gateway location and topologies

– With this value of p in mind, the probability that there are n nodes in the coverage area of a gateway can be computed as a binomial distribution (with N the total number of mobile nodes):

n N n

p p n N n g

−

−         = ) 1 ·( · ) (

Adaptive Gateway Algorithm (and IV)

5 10 15 20 25 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Nodes p=0.1 N=50 p=0.2 N=50 p=0.3 N=50

Cumulative density function, G(n) – The values of N1 and N2 must be chosen within the linear zone of G(n) – N2 will be the mean of MRA messages received – N1 is equal to the mean divided by 4 – But the values of N1 and N2 are dynamics and they change whenever a MRA message is sent by the gateway – The standard value of 2 seconds has been chosen for TMIN and the typical value of 20 seconds has been chose for TMAX

Performance evaluation (I)

2 to 5 m/s Speed Random Waypoint Model (RWP) Mobility pattern Ten sources 15 packets/s CBR Maximum speed: 2 m/s to 5 m/s Pause time: 10 s Mobility pattern Link layer AODV Local repair disabled Link layer detection enabled 802.11 RTS/CTS enabled Ad hoc protocol 250 m Transmission range

Simulation common parameters – Three different simulation settings are defined and used to validate the Adaptive Gateway Algorithm – The common parameters for the simulations have the following values:

Performance evaluation (II)

(625, 250) (1875, 250) (0, 0) (600, 600) (750, 150) Location gateway 100 75 50 Nodes 2500 x 500 m2 600 x 600 m2 1500 x 300 m2 Dimension SCENE III SCENE II SCENE I

Different simulation scenes – The simulations have been implemented in three different environments (node density, surface and gateway position):

• The Scene I corresponds to a rectangular area with the gateway

in the center of the topology

• The Scene II is a square area with two gateways, located in the opposite

corners of the square

• The Scene III is a wide rectangular area with two gateways

Performance evaluation (III)

– The simulations results of the AGW are compared with the MSC and the RMD algorithms (before explained) – A software module that includes the algorithm in the Global Connectivity support has been developed – And this module has been integrated into the Network Simulator, ns-2.29 on Linux – The algorithms have been tested in functions of these parameters:

• Packet loss rate (plr): defined as the ratio of the number of lost packets to the

total number of transmitted packets

• End-to-end delay (delay): it represents the average value of the time that the

received packets take to reach the destination

• Routing overhead normalized (ron): defined as the total number of control

packets divided by the total number of received packets – plr and delay values are the two most important parameters from the point of view of the userd. ron is important due to the need of having a measurement of the battery consumption in the mobile nodes

Performance evaluation (IV)

0.4539 0.4345 0.4138 0.3604 AGW 0.4827 0.4601 0.4451 0.3928 MSC 0.5163 0.4590 0.4357 0.3852 RMD ron 0.0582 0.0569 0.0474 0.0472 AGW 0.0629 0.0618 0.0548 0.0524 MSC 0.0650 0.0610 0.0519 0.0507 RMD plr 0.0800 0.0793 0.0729 0.0686 AGW 0.0858 0.0820 0.0813 0.0689 MSC 0.0862 0.0854 0.0817 0.0738 RMD delay 5 4 3 2 T Metric Maximum Speed

Scene I – A comparison between RMD and MSC algorithms points out that the RMD algorithm is better regarding to plr and ron, but not in the delay parameter – Moreover, with the proposed AGW algorithm, the best results are achieved

Performance evaluation (V)

0.2458 0.2235 0.1900 0.1548 AGW 0.2645 0.2389 0.2130 0.01678 MSC 0.2592 0.2362 0.2004 0.1659 RMD ron 0.0243 0.0241 0.0217 0.0153 AGW 0.0529 0.0463 0.0432 0.0278 MSC 0.0459 0.0411 0.0395 0.0179 RMD plr 0.1756 0.1371 0.1243 0.1069 AGW 0.2820 0.2214 0.2143 0.1252 MSC 0.3367 0.2383 0.2373 0.1745 RMD delay 5 4 3 2 T Metric Maximum Speed

Scene II – The results obtained for RMD and MSC algorithms are very similar because the proposed parameter in RMD algorithm will be always higher than the threshold due to the position of the gateways which, in case of the square area, are located in the opposite corners. The MSC algorithm obtains worse results than RMD – The proposed AGW algorithm presents the best results

Performance evaluation (and VI)

0.2017 0.1723 0.1515 0.1333 AGW 0.2252 0.2051 0.1645 0.1484 MSC 0.2185 0.1837 0.1599 0.1427 RMD ron 0.0075 0.0064 0.0063 0.0057 AGW 0.0129 0.0118 0.0109 0.0106 MSC 0.0120 0.0109 0.0106 0.0096 RMD plr 0.0274 0.0270 0.0257 0.0254 AGW 0.0346 0.0345 00.0305 0.0303 MSC 0.0377 0.0358 0.0325 0.0309 RMD delay 5 4 3 2 T Metric Maximum Speed

Scene III – For this new environment, the MSC algorithm is the worst of the three algorithms – The proposed AGW algorithm obtains, again, the best results (delay, plr and ron)

Conclusions (I)

– A new method to optimize the process that enables the internet connectivity in multi-hop ad hoc networks has been presented – The optimization minimizes the load control by choosing the T period taking into account the network connectivity – The AGW algorithm estimates the network connectivity counting the MRA messages received to get better end-to-end delay, packet loss rate and routing

• verhead that other proactive mechanisms

– The AGW algorithm also improves other schemes regardless of speed nodes, location gateways and mobitily patterns

AN ADAPTIVE GATEWAY DISCOVERY IN HYBRID MANETS

• F. D. Trujillo, A. Triviño, E. Casilari and A. Díaz-Estrella

Department of Electronic Technology University of Malaga

• A. J. Yuste

Department of Telecommunication Engineering University of Jaen