Performance of Different Node Densities

As we can observe, Spray and Wait, Epidemic and PROPHET have a shorter delivery delay in a smaller buffer size shown in Figure 24. Because the three algorithms perform a strategy of dropping message with oldest receiving time when the buffer is full, the condition of dropping message would frequently happen in a smaller buffer size. It would frequently cause that the message to be replaced by which with a shorter receiving time, that makes the average message delivery delay smaller.

As the buffer size becomes bigger as shown in Figure 25, Figure 26 and Figure 27, SFMS gradually outperforms the other algorithms and still maintain a differential performance among the four priorities of messages. Because when the buffer size becomes bigger, the other four algorithms would have bigger space to store more messages, and that would cause the average delivery delay longer. In other words, the message replaced by the one with a shorter receiving time would become less frequent.

4.3.2 Performance of Different Node Densities

We also analyze the effect of performance over different node densities. In DTNs, the delivery of message is based on the intermittent connection among nodes. Hence, if the number of the node increases, the chances of contacting a node would also increase theoretically. In other words, the chances that a message is carried by another node would increase. Because the increasingly chances of message relay, it needs an appropriate approach to choose the relay node, otherwise the delivery of message may become more and more inefficient. Therefore, in this section, we will analyze the five algorithms’ message delivery performance over different node densities. The performance results are shown in Figure 28, Figure 29 and Figure 30.

‧

Figure 28: Delivery ratio vs. Different node densities

Figure 29: Overhead ratio vs. Different node densities

0.35

Epidemic SAW Prophet UDM SFMS

0

Epidemic SAW Prophet UDM SFMS

‧

Figure 30: Delivery delay vs. Different node densities

Because the number of the node increases, the chances of contacting a node with a better delivery utility increase as well. Hence, the average delivery ratio of SFMS has a slightly enhancement in comparison with UDM. SFMS still control the overhead ratio well.

Moreover, the performance of delivery delay has a positive reflection as well. Because the strict restriction of message copies in Spray and Wait, it would not benefit from the increasing number of the node. Because Epidemic and PROPHET do not have a strict restriction on message copies, they could have a slightly enhancement on delivery ratio due to the widely increasing chances of delivering a message to another one. However, they will cause a direct proportion increasing to the overhead ratio.

In the end of this section, we will analyze the contention mechanism applied to SFMS in Figure 31 and Figure 32. The core concept of this mechanism is to let lower priority of message have a chance to contend with higher priority of message to acquire an earlier forwarding sequence.

Epidemic SAW Prophet UDM SFMS

‧

Figure 31: Delivery ratio gap vs. Different node densities

Figure 32: Overhead ratio gap vs. Different node densities

Due to the increasing chances of the node contact, a node would have more chances to forward a message to another node. It also indicates that the lower priority the message has,

0

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

44

the more chances to contend the forwarding sequence. This is very different from UDM, which has a fixed transmission sequence for different priorities of messages. In Figure 31, we can observe that SFMS has a smaller gap between the highest message priority and the lowest message priority when the number of node increases, and the same also happened in the gap of overhead ratio shown in Figure 32.

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

45

CHAPTER 5

Conclusions and Future Work

In this thesis, we propose a three-phase routing protocol for a message in the process of delivery. They are popularity spray phase, utility-based forwarding phase, and message delivery sequence phase, respectively. Firstly, popularity spray mechanism could distribute message to distinct N nodes more efficiently in a regular node mobility pattern than source spraying and binary spraying. Secondly, utility-based forwarding mechanism could consult the history of contact duration to further forward the message to another node with multi-copy to enhance the delivery performance when a message can not find its destination in the popularity spray phase. Thirdly, before actually transmitting the messages to the contact node, SFMS will let every message which is ready to be sent contend the forwarding sequence according to their priorities and costs that defined as the time aging from last contact of the destination. Through this scheduling mechanism, SFMS can not only forward message more accurate but also maintain a better resource allocation for all priorities of messages.

For further research, the calculation of predicting contact popularity and contact utility would be a topic that is worth to study. It is expected to be more accurate to reflect the future node state such as taking node moving speed into consideration and evaluates the effectiveness between popularity and hop counts for further adjusting the message spraying strategy.

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

46

REFERENCES

[1] J. Shen, S. Moh, I. Chung, “Routing Protocols in Delay Tolerant Networks: A Comparative Survey,” In ITC-CSCC, Chosun University, Korea, July 2008.

[2] E.P.C. Jones, P.A.S. Ward, “Routing Strategies for Delay-Tolerant Networks,” Submitted to Computer Communication Review, University of Waterloo, Canada, 2008

[3] R.H. Frenkiel, B.R. Badrinath, J. Borras, R.D. Yates, “The Infostations Challenge:

Balancing Cost and Ubiquity in Delivering Wireless Data,” In IEEE Personal Communications, vol. 7, no. 2, pp. 66-71, April 2000.

[4] L. Pelusi, A. Passarella, M. Conti, IIT-CNR, “Opportunistic Networking: Data Forwarding in Disconnected mobile ad hoc networks,” In IEEE Communications Magazine, Vol. 44, no.

11, pp. 134-141, November 2006.

[5] A. Vahdat and D. Becher, “Epidemic Routing for Partially-Connected Ad Hoc networks,”

In Technical Report CS-2000-06, Duke University, July 2000.

[6] A. Demers, D. Greene, C. Houser, W. Irish, J. Larson, S. Shenker, H. Sturgis, D. Swinehart, and D. Terry, “Epidemic Algorithms for Replicated Database Maintenance,” In SIGOPS Operating Systems Review, vol. 22, pp. 8-32, January 1988.

[7] T. Spyropoulos, K. Psounis, C.S. Raqhavendra, “Spray and Wait: An Efficient Routing Scheme for Intermittently Connected Mobile Networks,” In Proc. SIGCOMM, EE, USC, USA, August 2005.

[8] A. Lindgren, A. Doria, and O. Schel’en, “Probabilistic Routing in Intermittently Connected Networks,” In Proc. SIGMOBILE, vol. 7-3, LUT, Sweden, July 2003.

‧

[9] J. LeBrun, C.N. Chuah, D. Ghosal, M. Zhang, “Knowledge-Based Opportunistic Forwarding in Vehicular Wireless Ad Hoc Networks,” In Proceedings of IEEE Vehicular Technology Conference (VTC), vol. 4, pp. 2289-2293, University of California, Davis, USA, May 2005.

[10] T. Umedu, H. Urabe, J. Tsukamoto, K. Sato, T. Higashino, “A MANET Protocol for Information Gathering from Disaster Victims,” In PERCOMW, Osaka University, Japan, March 2006.

[11] X. Pallot, L.E. Miller, “IMPLEMENTING MESSAGE PRIORITY POLICIES OVER AN 802.11 BASED MOBILE AD HOC NETWORK,” In MILCOM, NIST, USA, 2001.

[12] C. Suthaputchakun, A. Ganz, “Military Inter-Vehicle Communication with Message Priority using IEEE 802.11e,” In MILCOM, University of Massachusetts Amherst, USA, October 2006.

[13] H. Gong, J.W. Kim, “A Prioritization-Based Application-Oriented Broadcast Protocol for Delay-Tolerant Networks,” In WCNC, GIST, Korea, 2009.

[14] Z. Li, H. Shen, “Utility-based Distributed Routing in Intermittently Connected Networks,”

In ICPP, University of Arkansas, USA, September 2008.

[15] Z. Li, H. Shen, “SEDUM: Exploiting Social Networks in Utility-Based Distributed Routing for DTNs,” In IEEE Transactions on Computers, Vol. 62, no 1, pp. 83-97, Clemson University, USA, January 2013.

[16] H. Dubois-Ferriere, M. Grossglauser, M. Vetterli, “Age Matters: Efficient Route Discovery in Mobile Ad Hoc Networks Using Encounter Ages,” In MobiHoc, EPFL, Switzerland, June 2003.

[17] I. Joe, K. Sang-Bo, “A Message Priority Routing Protocol for Delay Tolerant Networks (DTN) in Disaster Areas,” In FGIT, Hanyang University, Korea, December 2010.

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

48

[18] R. Ramanathan, R. Hansen, P. Basu, R. Rosales-Hain, R. Krishnan, “Prioritized Epidemic Routing for Opportunistic Networks,” In MobiOpp, BBN Technologies, USA, June 2007.

[19] Z. Li, L. Sun, E.C. Ifeachor, “Adaptive Multi-Copy Routing for Intermittently Connected Mobile Ad Hoc Networks,” In GLOBECOM, University of Plymouth, UK, December 2006.

[20] T. Spyropoulos, K. Psounis, C.S. Raghavendra, “Spray and Focus: Efficient Mobility-Assisted Routing for Heterogeneous and Correlated Mobility,” In PerCom, March 2007.

[21] A. Keranen, J. Ott, T. Karkkainen, “The ONE Simulator for DTN Protocol Evaluation,”

In SIMUTools, Helsinki University, Finland, March 2009.

[22] Haris, Abdullah, “A DTN Study: Analysis of Implementations and Tools,” In IMM-MSC-2010-65, Technical University of Denmark, Denmark, 2010.

[23] M. McNett, G.M. Voelker, “Acm Mobile Computing and Communication Review,” In Access and mobility of wireless pda users, 2007.

[24] Nielsen, http://www.nielsen.com/us/en.html.

[25] eMarketer, http://www.emarketer.com/.