Proceedings of IC-BNMT2009

FORMAL TAXONOMY RESEARCH ON OPPORTUNISTIC NETWORKS Zhou Qiang1, Ying Jing1,2, Wu Minghui1,2 1 College of Computer Science, Zhejiang University 2 Zhejiang University City College [email protected], [email protected], [email protected]

Abstract Opportunistic networks have complex construction and are multi-formed. In order to standardize the research on opportunistic networks, a formal taxonomy was proposed, which formally defined the geographical preferences, movement terminals and special transmit nodes of opportunistic networks. Then the opportunistic networks were classified according to the taxonomy. This taxonomy provided an open platform for the other research fields of opportunistic networks. Keywords: taxonomy

opportunistic

networks;

formal;

whales). There are applications for pocket wireless devices such as PSN[6], and for in-vehicle devices networks such as CarTel[7]. And there is a variety of other applications[2]. Therefore, opportunistic network is a valuable and potential research field. Mobility models play an important role in the research of mobile networks. The current research on the opportunistic networks is usually "bottomup" – in other words, it’s based on a particular mobility model or realistic application, and did not form a unified specification. In order to standardize a "top-down" research method on opportunistic networks, in this paper we will analyze the current research on opportunistic networks and then propose a formal taxonomy.

1 Introduction

2 Related work

In the classical fields of network communication, connected topology is a basic request. It means, when the messages transmit from the source node to the destination node, a data path exists between them. However, as the wireless networks are extensively used in more and more fields, the structure of networks has become much complex and unstable. In some realistic applications of the network, the information transmission delay may be far more than the classic network latency, the network connectivity can not be guaranteed. Under certain circumstances, the network nodes themselves are ambulant, which has led to such a phenomenon: the network itself is not connected, but the movements of the nodes have the opportunity to meet other nodes. Then the messages may be transmitted from the source node to the destination node by taking advantage of the meeting opportunities. Therefore, the concept of “opportunistic networks” has been founded[1]: the nodes of opportunistic network exploit the meeting opportunities with others to transmit messages while the topology of the network is disconnected.

In opportunistic networks, the messages are transmitted when the nodes opportunistically meet the other peer. Therefore, the characteristics of node movement will greatly affect the message transmission, which made the research on mobility models an important point of the study. As the structure of opportunistic networks is unstable and the movement characteristics differ from each other, there are diversified mobility models. In general, we have classic mobility models and realistic mobility models.

There are numerous applications of opportunistic networks because of the unstable and disconnected network structures of the realistic self-organizing ad hoc networks. There are applications for tracking wild animals such as ZebraNet[4] (for tracking Zebras) and SWIM[5] (for tracking ___________________________________ 978-1-4244-4591-2/09/$25.00 ©2009 IEEE 

Classic mobility models described a chaotic situation of networks, in which the nodes are moving disorderly and aimlessly. In this case, the mobility model is IID (Identical and Independently Distributed)[3]. It means that the trajectories of each node will not be impacted by other nodes, and each node moves non-differently from others. RWP (Random Waypoint), RD (Random Direction) and RW (Random Walk) are examples of classic mobility models. As the nodes’ movements are completely random in such classic models, the circumstances are relatively simple and provided an easier way to study the routing protocols of the opportunistic networks. However, the realistic opportunistic network nodes’ movements are usually not completely random, so the classic model is often divorced from the realistic situations

and cannot portray the realistic opportunistic networks well. Realistic mobility models described the actual applications of opportunistic networks. In practice, the network nodes can be the wireless equipments carried by people, vehicles or even wild animals. In order to portray these realistic opportunistic networks, researchers put forward statistics-based models and community-based models. Such models may describe the characteristics of the nodes’ movements via statistical methods instead of the random-moving suppose, and community characteristics can often be obtained in such models. Realistic mobility models are usually more complex than the classic ones, but they are able to portray the actual opportunistic networks.

probability

pk exceeds the threshold D .

Geographic preference of nodes can not be described correctly if the value of threshold D is too large or too small. For the whole opportunistic network N, the set of "Frequently Reached Location" of all the nodes is defined as FN : Definition 1.2:

FN

*F

i

1di  | N |

There are FN

) for classic mobility models, and

FN z ) for realistic mobility models. 3.2 Movement terminals of nodes

Both classic and realistic mobility models described above are based on specific opportunistic networks. In other words, they are bottom-up models.

3 Formal taxonomy We are looking for a unified specification of opportunistic networks research, which needs a topdown analysis. This section discusses the significant features of the opportunistic networks, and then proposes a formal taxonomy. 3.1 Geographic preferences of nodes Classic mobility models such as RWP and RD assume that the nodes are moving randomly. So the arriving probabilities at different locations of different nodes are equal. That is to say, there is no preferred location of any node. However, the nodes’ movements are not completely random in realistic mobility models. The movements might be based on community. Some nodes may tend to move towards a particular destination. Some nodes’ movements may be restricted within a particular area. Some nodes may prefer to meet specific nodes than others. So actually the nodes usually have geographic preferences. Now we give out the definitions. For a node i in opportunistic network N, the set of "Frequently Reached Location" is defined as Fi : Definition 1.1:

Fi

{( Lk , pk ) pk ! D } 

( Lk , pk ) is a two-tuple. Lk is a frequently reached

location of node i. pk is the arriving probability at

Lk of node i in unit time. D is a predetermined threshold. We consider the node i to have "Frequently Reached Location" Lk if the

The movements of nodes in opportunistic networks usually have time or space limitations. A node reaches its time terminal when the node can no longer transmit messages even if it is still moving. For instance, when the wireless device battery of a node has run out, it will stop transmitting messages and be considered to reach its time terminal. On the other hand, a node of an opportunistic network reaches its space terminal when the node leaves the network geographically. Taking an opportunistic network consist of the wireless equipments carried by vehicles in a highway for example: when a car node exits the highway, it leaves the current opportunistic network geographically, which means it has reached its space terminal of the network. Moreover, the node might go on moving and join a new opportunistic network after it leaves the current one, but it does reach the current space terminal because we do not concern about the new one. We call both time and space terminals "Terminal of Movement". We can estimate whether a node has reached its terminal by the formula defined as below: Definition 2.1:

ToM (i ) ToM ( Li , TP i ) 1 

| Li  TLi | TPi ˜ D max T max

TLi is the location of space terminal of node i in opportunistic network. Li is the current location of node i. Dmax is the length of geographic diameter of the opportunistic network. TPi is the remaining power supply time of node i. Tmax is the maximum power supply time of node i. We can determine that node i has reached its terminal when ToM (i ) 1 . In particular, the movements of the network nodes might have no time limitations – for example, the



power supply is infinite – then we have TPi and Definition2.1 has become: Definition 2.2:

ToM (i ) ToM ( Li ) 1 

Tmax ,

| Li  TLi |  D max

On the other hand, the movements of the nodes might have no space limitations – for example, the nodes will never leave the network – then we transform the Definition2.1 into: Definition 2.3: TPi ToM (i ) ToM (TPi ) 1  T max If the terminals of movements exist in an opportunistic network, there is ToM (i )  [0,1] . We can infer that ToM (i) { 0 if the nodes in opportunistic network do not have movement terminals. 3.3 Special nodes

The fixed nodes are a kind of special nodes in opportunistic networks. The fixed nodes will no change there locations, and they are usually well equipped: they may have no limitation of power supply, and have high performance of message transmitting and so on. For example, the fixed nodes in an office building can provide the data storage and transmitting services for the portable wireless devices carried by people. A fixed node in the network can be a landmark for the other moving nodes. In this paper, we call the fixed nodes "base station", as the definition below: Definition 3.1:

{(bk , Lk , ck ) k

TN

TN is the set of all transporter nodes t k in opportunistic network N. ck is the storage capacity of t k . There is TN ) for the opportunistic network N which has no transporter node. 3.4 Classification of opportunistic networks We can determine FN , Tom(i ) , B N and TN by the definitions above. Then we have a classification of opportunistic networks according to these characteristics as Figure 1:

characteristic FN

Description )

FN FN z )

ToM (i)  [0,1]

ToM (i)

ToM (i ) { 0

1,2,3...}

B N is the set of all base stations bk in opportunistic network N. Lk is the location of base station bk . ck is the storage capacity of bk .

for the opportunistic network N which has base stations. In particular, if | B N | | N | , then all the nodes in the network are fixed, which means the opportunistic network N has degenerated to a common ad hoc network. A second kind of special nodes in opportunistic networks are "transporter nodes", which are defined as "ferry" in some other literatures. These nodes can take the initiative to move toward specific nodes which they need to meet. These nodes have

.realistic opportunistic networks .the nodes' movements have preferences .often community based .the nodes' movements have time or space limitations .nodes will reach the terminal at specific time or location .the nodes' movements have no time or space limitations .nodes will never leave the network

)

BN z )

.no base station exists in the opportunistic networks

)

.transporter nodes exist in the opportunistic networks .transporter nodes can take the initiative to move toward specific nodes

TN z )

no transporter node exists in the opportunistic networks

BN

) for the opportunistic network N which has no base station, while there is BN z ) 

.classic opportunistic networks .the nodes' movements are IID

.base stations exist in the opportunistic networks .base stations have higher performance than other nodes

BN

There is BN



{(t k , ck ) k 1,2,3...}

Classification of Opportunistic Networks

In opportunistic networks there may be special nodes which have different behaviors than the common nodes.

BN

the knowledge about the network, so when they need to transmit a message to a specific node, they will move to look for it. They will have higher probability to meet the destination nodes than the common ones. Definition 3.2:

TN

TN

Figure 1. classification of opportunistic networks The formal taxonomy we proposed above is open. Other characteristics can be added to this taxonomy.

According to this taxonomy, we can classify an opportunistic network first, and then begin to study the problems such as routing protocols using the knowledge of the classification.

4 Conclusions Now the researches on opportunistic networks are usually based on a specific model or application. In order to found a unified research specification, we formally define the characteristics about the geographic preferences, movement terminals and special transmit nodes of the opportunistic networks in this paper. And then a taxonomy is proposed, which provides an open platform and top-down method for the opportunistic networks studies. The platform can be extended by formally adding other characteristics to it. Top-down studies on opportunistic networks such as the routing protocols can be proceeding by taking advantage of this taxonomy.

References [1] Pelusi L, Passarella A, Conti M. Opportunistic networking: data forwarding in disconnected mobile ad hoc networks[J]. Communications Magazine, 2006,44(11):134141. [2] Xiong Yongping, Sun Limin, Niu Jianwei, et al. Opportunistic Networks[J]. Journal of Software, Vol.20, No.1, January 2009:124137 [3] Spyropoulos T, Psounis K, Raghavendra C S. Performance Analysis of Mobility-assisted Routing[C]. MobiHoc’06, May 22–25, 2006, Florence, Italy [4] Juang P, Oki H, Wang Yong, et al. Energyefficient computing for wildlife tracking: Design tradeoffs and early experiences with ZebraNet[C]. Proceedings of the 10th Int’l Conf. on Architectural Support for Programming Languages and Operating Systems, New York: ACM, 2002. 96107. [5] Small T, Hass Z. The shared wireless infostation model: A new ad hoc networking paradigm (where there is a whale, there is a way)[C]. Proceedings of the 4th ACM Int’l Symp. on Mobile Ad Hoc Networking,Annapolis: ACM, 2003. 233244. [6] Pan Hui, Chaintreau A, Scott J, et al. Pocket switched networks and human mobility in conference environments[C]. Proceedings of the 2005 ACM SIGCOMM Workshop on Delay-Tolerant Networking. Philadelphia: ACM. 2005.244251. [7] Hull B, Bychkovsky V, Zhang Y,et al.CarTel: A distributed mobile sensor computing system[C]. Proceedings of the 4th Int’l Conf.

on Embedded Networked Sensor Systems. Boulder: ACM, 2006. 125138. [8] Song Libo, Kotz D F. Evaluating Opportunistic Routing Protocols with Large Realistic Contact Traces[C]. CHANTS’07, September 14, 2007, Montréal, Québec, Canada. [9] Boldrini C, Conti M, Passarella A. ContentPlace: Social-aware Data Dissemination in Opportunistic Networks[C]. MSWiM’08, October 27–31, 2008, Vancouver, BC, Canada. [10] Abdulla M, Simon R. Characteristics of Common Mobility Models for Opportunistic Networks[C]. PM2HW2N’07, October 26, 2007, Chania, Crete Island, Greece. [11] Boldrini C, Conti M, Passarella A. Modelling Data Dissemination in Opportunistic Networks[C]. CHANTS’08, September 15, 2008, San Francisco, California, USA. [12] Collins M P,Shimeall T J, Faber S, et al. Using Uncleanliness to Predict Future Botnet Addresses[C]. IMC’07, October 24-26, 2007, San Diego, California, USA. [13] Pan Hui, Crowcroft J. How Small Labels create Big Improvements[C]. Proceedings of the 2006 ACM CoNEXT conference, Lisboa, Portugal,2006: Article No. 34 [14] Campbell A, Eisenman S, Lane N, et al. People-centric urban sensing[C]. IEEE Wireless Internet Conference, August 2006. [15] Henderson T, Kotz D, Abyzov I. The changing usage of a mature campus-wide wireless network[C]. Proceedings of the 10th Annual International Conference on Mobile Computing and Networking (MobiCom), pages 187–201, September 2004.