Other application

The method of RTS/CTS detection can be also applied to other wireless handover approaches. In this section, some examples is illustrated.

3.7.1 HAWAII

With HAWAII, every router keeps a map from its output port to a mobile node, so they all know how to forward packets for mobile nodes. When a mobile node handovers to a new FA, it sends mapping update to both the gateway router and the old FA. Every router along the path will update its mapping to that mobile node.

The old FA helps the mobile node to buffer packets, but every packet for the mobile node will be buffered regardless of being received successfully by the mobile node. The buffer space will be wasted.

If we can use RTS/CTS to detect the handover of mobile nodes, we can use

buffer space more efficiently. In other words, the FA buffer the packets only when the mobile node is expected to move out of the current coverage area.

3.7.2 Hierarchical structure

The packets loss problem presented in last section can be solved by RTS/CTS detection and packets buffering/forwarding. Therefore packets will be never lost in the old FA.

3.7.3 fast handover

Fast handover only works when a mobile node can listen to more than one FA simultaneously. If we can add our method to it , its application range can be extended.

Chapter 4 Performance evaluation

To verify that the new model proposed in Chapter3, we perform simulation with network simulator ns2 version 2.26. There is a mobile IP extension included in this version. We modify some code for our model, and record the result. The result is satisfying and described as follows.

4.1 Simulation I

In this simulation, we want to see whether the speedy handover works well or not. We can observe the performance of wireless handover with packet buffering and forwarding. To avoid the complex TCP mechanism such as retransmission, we use UDP as the test traffic. By checking the sequence number, the number of packets loss can be detected.

4.1.1 Network topology

We would like to evaluate the performance during handover of a mobile node, so we built up a network topology which has one mobile node (MN), one corresponding node (CN), two foreign agent (FA), and one home agent (HA) as shown in Figure 4.1.

Figure 4.1 The topology of simulation I.

In Figure 4.1 each wired link has 5Mb bandwidth, 2ms delay time and drop tail queue policy. CN and router are wired device and only support wired routing. HA, FA1 and FA2 are wireless agents including access router with access point together.

They have wire and wireless interface and routing policy. MN (mobile node) is a wireless device with wireless routing only.

The radio propagation model setup in ns2 is as follows:

Distance = 550m

Propagation model: TwoRayGround Selected parameters:

Transmit power: 0.281838 Frequency: 9.14e+08 Transmit antenna gain: 1 Receive antenna gain: 1

System loss: 1

Transmit antenna height: 1.5 Receive antenna height: 1.5

Receiving threshold RXThresh_ is: 1.55924e-11

The communication range of all nodes with wireless interface is 550m. The distance between FA1 and FA2 is 856m and the overlap of communication range is about 244m.

The simulation starts at 0s and ends at 180s. A UDP sender (CN) starts to send packets at 100s until the end of simulation, and the mobile node begins to move from the FA1 to the FA2 at the same time with speed 20m/s.

The CN is the sender of UDP traffic, which is setup as follows:

$cbr set type_ CBR

$cbr set packet_size_ 1000

$cbr set rate_ 1mb

$cbr set random_ false

The packet size is 1000bytes and the sending rate is 1 Mbps. The receiver of UDP is a mobile node. We give each packet a sequence number, by checking the packet sequence numbers the mobile node received, we can observe which packet is delivered successfully and which one gets lost.

We run the simulation with existing method and speedy handover, the result is described in the next section.

4.1.2 Simulation result

The handover begins at about 127.46s, so we show the result from 127s to 131.5s as follows:

Figure 4.2 Packet sequence number the mobile node received in the original structure in simulation I.

Figure 4.3 Packet sequence numbers the mobile node receives with speedy handover in simulation I

In Figure 4.2, during handover period when the mobile node is unable to receive any packet from UDP sender, (CN) every packet get lost because no mechanism is available to buffer and forward the packets for the mobile node. Obviously the packet loss rate is very high and the performance is degraded seriously.

In Figure 4.3 the speedy handover is used. The old FA buffers the packets, and forwards them as soon as the new FA knows where to deliver the packets to the mobile node which just completed the routing update. The mobile node can receive packets before handover procedure finishes. The handover procedure is complete at around 129.15s and the mobile receives forwarded packets at around 128.66s, so we shorten the time during which the mobile node can not receive any packet. After handover procedure finishes, the new packet stream toward the mobile node make the packet arrival out of order. It is the cause of the heavy line in Figure 4.3.

Figure 4.4 Packets loss rate.

Observing Figure 4.4, we can find clearly that packets lose rate with speedy handover is much lower than with original structure. With original structure the rate is 55.41%, and with speedy handover the rate is 1.61%. By bufferring and forwarding packets loss is only caused by out of space of interface queue and random loss. That is why lower packets loss rate is achieved.

When we have a multi-media service, low packet lose rate is an important requirement for quality of service. According to the Figure 4.4, our proposed method is suitable for multi-media service.

Figure 4.5 Handover time of simulation I.

Figure 4.5 shows that the time from the start to the end of handover period while the mobile node can not receive any packet. It is clear that in speedy handover the handover time can be shorten..

The total handover time is about 1.576439 seconds. Since the old FA in speedy handover forwards packets before the handover finishes, the mobile node can receive packets earlier than in the original structure. On the other hand, the time spent in speedy handover is only 1.19214 seconds, that is, it spends 384.299 ms less.

For multi-media service, the shorter handover time means a smoother quality of service. For example, the frames of movie can be played out much smoothly and the client can experience a higher quality.

Figure 4.6 Buffer utilization of simulation I.

The buffer utilization is another issue we have to address, thus we observe the buffer utilization during simulation. The result is shown in Figure 4.6. The mobile node detaches from the old FA at about 127.5s, so a burst of packets is queued in buffer after 127s until “request to forwarding” message is received, after that the packets are removed to the output interface queue.

Although there are a little detection failure before handover and packets are added to the buffer, through further detection with RTS/CTS the buffer can be cleared immediately. The detection failure therefore does not cause a big problem, and we never waste too much buffer due to this reason.

4.2 Simulation II

In this simulation we would like to evaluate the performance when the coverage areas of the two FAs do not overlap. In this situation the fast handover does not work.

4.2.1 Network topology

Performance during the handover of a mobile node is what we would like to know, so we build up a network topology consisting of one mobile node (MN), one corresponding node (CN), two foreign agent (FA), and one home agent (HA) as that in simulation I in the Figure 4.1. However, the radio propagation model is different.

Distance = 400m

Receiving threshold RXThresh_ is: 5.57346e-11

The communication signal range of all nodes with wireless interface is 400m.

The distance between FA1 and FA2 is 856m and the overlapped range of communication signal is none. The distance between the margins of two

communication range is about 56 m as shown in Figure 4.7:

Figure 4.7 Communication signal range of simulation II.

The simulation starts at 0s and ends at 180s. An UDP sender (CN) starts to send packets at 100s until the end of the simulation, and a mobile node begins to move from FA1 to FA2 at the same time with speed 20m/s.

The CN is a sender of UDP traffic, which is setup as follows:

$cbr set type_ CBR

$cbr set packet_size_ 1000

$cbr set rate_ 1mb

$cbr set random_ false

The packet size is 1000bytes and the sending rate is 1 Mbps. The receiver of UDP is the mobile node. We give each packet a sequence number. By checking the packet sequence numbers the mobile node has received, we are able to determine which packet is delivered and which is lost.

We measure the performance with the sequence numbers of packets received by the mobile node. The result is described in the next section.

4.2.2 Simulation result

Every time we get a packet we record its sequence number and the time it is received by the mobile node. We record the result using the original structure as well as in the speedy handover in Figure 4.8.

Figure 4.8 The result of the original method in simulation II.

We compare the result with that in simulation I. Because of the longer distance of two FAs, the handover time is longer, and more packets get lost. To show the effectiveness of our new method, we show the result of speedy handover in Figure 4.9.

Figure 4.9 The result of speedy handover in simulation II.

In Figure 4.8 and Figure 4.9, it is clear that speedy handover does work when the communication ranges of two FAs do not overlap. In Figure 4.8, every packet get lost during handover, which in speedy handover this problem is improved. The packets are forwarded by the old FA, and are received by the mobile node successfully. Besides, we also can find that the time period during which the mobile node can not receive any packet is shortened significantly. The mobile node is able to receive packets before the whole handover processes finishes.

Figure 4.10 The handover time of simulation II.

The total handover time period during which a mobile is unable to receive any packet has been shortened in our proposed speedy handover as shown in Figure 4.10.

We also demonstrate that speedy handover also works well when the communication signal ranges of two FAs do not overlap.

On the original structure the elapsed time a mobile node can not receive packets is about 12.573266 s. Because of shorter communication range, the handover time is much longer than that in simulation I.

With speedy handover the time a mobile node can not receive packets is about 7.790617 s. It is much shorter (38.04% less) than in the original structure.

Just as that in simulation I, there is a problem regarding out-of-order packets. For on-demand video service, we can set a buffer in the mobile node to solve the problem.

When the mobile node receives out-of-order packets, it can put them into the buffer,

and wait other packets for a short period. When packets become in order, the mobile node plays out the video frame of that packets. The user also can get a smooth video.

However, for VoIP service, out-of-order packets problem is serious, because VoIP is interactive service and we can not buffer packet too long in the mobile node.

Therefore we will consider this problem of out-of-order packets in the future work.

Chapter 5 Conclusion

In this thesis, we propose a new approach to detect the handover activity of mobile nodes and buffer packets at a suitable time without wasting buffer space. We use short RTS/CTS messages for detection. The old FA buffers packets and forwards them as soon as possible. We successfully avoid packets loss during handover and allows the mobile node to receive data packets before handover completes. For multi-media service, the delay time can be shortened, so the quality can be improved.

With simulation result, the UDP packet loss rate during handover decreases significantly from 55.41% to 1.61% (speedy handover). The time period mobile nodes are unable to receive packets is also shortened from 1.576439s to 1.19214s which is 24.38% , based on our method. Thus we can be sure that our proposed method works greatly and improves delay time for multi-media service. Besides, we only buffer the packets when the mobile node has left. In other words, we never waste any buffer resources in either FA or AR, although there may be a little detection failures before handover.

When the communication signal ranges of two FAs do not overlap, speedy handover can still work well. In the simulation II, it is clear that the mobile node get more packets in 180s than the one with original method. The packet loss is decreased successfully and the handover time is shortened, too.

There are some problems remained to be solved. With simulation II result, out-of-order packets may be a problem. For on-demand video service, if we set a video buffer in the mobile node, the problem is not so serious, but for VoIP service, it

is a big problem. The voice delay is not tolerated when we speak to somebody through VOIP. It is a good subject for future studies.

The method proposed in this thesis is designed under the DCF structure. We may use the similar method under the PCF structure, too. We can let the polling messages work as RTS/CTS messages. This method under PCF structure also remains to be a future research.

Reference

[1] G. Dommety et al., “Fast Handovers for Mobile IPv6,”IETF draft, draft-ietf-mobileip-fast-mipv6-03.txt, July 2001.

[2] IEEE 802.11b, IEEE Std. 802.11-1999..

[3] H. Soliman et al., “Hierarchical MIPv6 Mobility Management,”IETF draft, draft-ietf-mobileip-hmipv6-05.txt, July 2001..

[4] R. Ramjee, T. La Porta, S. Thuel, K. Varadhan and L. Salgarelli, IP micro-mobility support using HAWAII, Internet draft, draft-ietfmobileip-hawaii-00.txt (June 1999)..

[5] The Network Simulator – ns-2, http://www.isi.edu/nsnam/ns.

[6] Nicolas Montavont and Thomas Noël, “Handover Management for

Mobile Nodes in IPv6 Networks,” IEEE Communications Magazine, August 2002.

[7] Marc Torrent-Moreno, Xavier P`erez-Costa, Sebasti`a Sallent-Ribes, “A Performance Study of Fast Handovers for Mobile IPv6”, Proceedings of the 28th Annual IEEE International Conference on Local Computer Networks (LCN’03).

[8] ANDREW T. CAMPBELL, JAVIER GOMEZ, SANGHYO KIM, AND CHIEH-YIH WAN, “COMPARISON OF IP MICROMOBILITY PROTOCOLS”, IEEE Wireless Communications • Feburary 2002.

[9] Janise McNair, Ian F. Akyildiz, Michael D. Bender, “Handoffs for Real-Time Traffic in Mobile IP Version 6 Networks”, 2001 IEEE.

[10] Guenkova-Luy, T.; Kassler, A.J.; Mandato, D, “End-to-End Quality-of-Service Coordination for Mobile Multimedia Applications”, Selected Areas in

Communications, IEEE Journal on , Volume: 22 , Issue: 5 , June 2004.

[11] Helmy, A.A.-G.; Jaseemuddin, M.; Bhaskara, G, “Multicast-based mobility: a novel architecture for efficient micromobility”, Selected Areas in Communications, IEEE Journal on , Volume: 22 , Issue: 4 , May 2004.

[12] Seok Joo Koh; Moon Jeong Chang; Meejeong Lee, “mSCTP for soft handover in transport layer”, Communications Letters, IEEE , Volume: 8 , Issue: 3 , March 2004 [13] Marsan, M.A.; Ginella, G.; Maglione, R.; Meo, M.. “Performance analysis of hierarchical cellular networks with generally distributed call holding times and dwell times” , Wireless Communications, IEEE Transactions on , Volume: 3 , Issue: 1 , Jan.

2004