International Journal of Electrical & Computer Sciences IJECS-IJENS Vol: 10 No: 01

87

Investigation of Packet Loss Patterns in Audio/Video Content Distribution over RTP in Wireless Networks Wajahat Abbas†, Misbah Zareen†† , Asim Shahzad††† , Lala Rukh†††† [email protected], [email protected], [email protected], [email protected] † †† ††† ††††

Faculty of Telecom. & Information Engg. , University of Engineering and Technology Taxila, Pakistan

Summary Real Time Multimedia brings a promise of providing real time streaming audio/video data over wireless networks. However, there is a problem associated with multimedia transmission over wireless networks i.e. packet loss. To achieve desired Quality of Service (QoS) we need to understand packet loss patterns in wireless networks. This paper holds the study of packet loss patterns in multimedia transmission of congested networks over RTP. We use NS2 as simulation tool and wavelet analysis is used to explore the packet loss behavior over wireless networks.

Key words: Network Simulator (NS-2), RTP/RTCP protocol, Multimedia Transmission, QoS

1. Introduction Multimedia applications require high QoS to cope with packet loss, jitter and delay. These parameters have strong impact on multimedia in terms of wireless networks and they adversely affect the throughput. Loss probability itself is not enough to explore the behavior of packet loss, Packet loss patterns need to be studied to investigate QoS parameters[1].We use NS2 simulator to capture the packet loss traces of multimedia data over RTP. Wavelet analysis is used to study the behavior in terms of time scale.Section2 describes the topology and traffic sources in detail.Section3,8 represent simulation results using NS2 simulator. Section 4 focuses on wavelet analysis and its performance [2].

Parameters Link speed between router and sources Link speed between sink and router Protocol is used to transfer the packets over wireless network FIFO buffer Buffer size Delay

Value 20 Mbps 55.746 RTP 300bytes 128 to 1,024 5 to 46ms

Delays below 15ms lie within the range of latencies that we encounter in interactive multimedia. Figure 1 shows the nam window of network topology. This figure shows a router along with 10 sources and a sink.

2. Simulation Scenario Network topology we assume here consists of 11 sources that generate multimedia traffic, a buffer and a common sink node. For our scenario the parameters are following:

Figure1: Nodes connected to sink node 1

International Journal of Electrical & Computer Sciences IJECS-IJENS Vol: 10 No: 01

88

Contribution of lost packets

Packet Loss 15 10 Series1

5 0 1

2

3

4

5

6

Lengthof brustspackets Figure 2: Packet loss graph for node 2, 3,4,5,6

Figure 3: Lost packets contribution of various lengths from lost brusts to overall number of lost packets.

3. Packet Loss

5. Packet Loss Generator

For simulation scenario we follow the topology used in section2.NS2 simulator is used to record packet loss traces. Experiments conducted using NS2 indicate that packet loss is highly brusted. When a previous packet is successfully received packet lost burst starts with a lost packet. For instance if packet 2, 5, 7 have been successfully received and packets 3, 4, 6 have been lost then burst starts a packet lost 2 and ends at packet lost 5.

This particular module is developed to examine packet losses percentage of RTP multimedia streams. We have shown above that to simulate packet loss individual packet loss models could be used.

b

Figure 2 shows a plot that depicts various graphs for different buffer sizes and for different number of nodes. Points on the graph indicate the time when packet loss occurs.

4. Packet Loss Wavelet Analysis

Investigation

Using

N 1-b

1-a

Correlation structure of multimedia traffic is complex and has long range dependence LRD [3].To analyze the properties of multimedia traffic Wavelet analysis is a very useful tool. To analyze the packet loss behavior and examine LRD properties of loss burst. Wavelet Analysis is used for this purpose.

L a

Figure 4: Packet loss simulation (Gilbert model)

L is a loss state and N is a state with no losses. a and b denotes the probabilities of being in one of the states i.e. either L or N respectively. Now, the matrix

P=

b 1-a

1-b a

International Journal of Electrical & Computer Sciences IJECS-IJENS Vol: 10 No: 01

represents the transition probabilities. Two parameters completely define the whole model, the probability of being in L (unconditional loss probability) pul and the transition probability from state N to state L. Where pcl is the conditional probability of loss: 1-pcl

B=1-pul

1-pul

1-a =

1-pul 1-pul

89

Spectral density is the equivalent property of LRD function’s (lambda) of Y(t). F(lambda)~ bf | lambda | 1-2M lambda 0 bf>0 and 0.5 0 and 0.5