You are on page 1of 7

Interest-Aware Content Distribution Protocol for Mobile

Disruption-Tolerant Networks

Arezu Moghadam and Henning Schulzrinne

Columbia University, New York, NY

Abstract multicasting in mobile DTNs try to extend the classical

multicast model to disruption-tolerant networks. They
Most of the routing protocols for mobile disruption- define new multicast semantics to overcome the intrinsic
tolerant networks (DTNs), are designed to route data network partitioning in mobile DTNs [21,7]. However,
from a single source to a single destination. Many these proposed architectures require a global knowledge
real-world applications, however, need to transfer data of the multicast group memberships of nodes and the
to a group of recipients rather than individuals. Pro- network topology [21]. Therefore, implementing these
posed architectures for multicast communication in architectures is infeasible in mobile DTNs which suffer
mobile DTNs rely on a global knowledge of multi- from a frequently changing topology and lack of infras-
cast group memberships that makes them infeasible for tructure to track group memberships.
infrastructure-less networks. Our interest-aware data Many real-world scenarios benefit from an intelli-
distribution protocol distributes data among a commu- gent content distribution algorithm that is able to learn
nity of mobile users in DTNs based on their inter- users' behaviors and their interests over time. This al-
ests. Our interest-aware algorithm automatically learns gorithm should be able to categorize mobile users based
users' interests from the history oftheir cached data and on their interests and distribute data among its interested
provides them with their preferred data content. users. In this paper we propose such an algorithm that
We implemented this protocol in the ONE simulator automatically learns users' behaviors and extracts their
and evaluated its performance using reality-mining mo- preferences to classify them into appropriate categories.
bile traces. Our simulations show that our protocol de- Our algorithm extracts users' interest-vectors from their
livers 30% more relevant data to mobile users' interests cached data content and uses these vectors to route spe-
than epidemic routing. Furthermore, this superior per- cific data to its interested recipients. When two mobile
formance is achieved with 35% lower distribution ofir- users meet, they exchange these interest-vectors to cal-
relevant data content among mobile users. culate the other user's interest in the data content they
are trying to distribute. The data is then transferred to
the other user only if there is enough correlation between
1 Introduction the user's interests and the data content. Therefore, un-
like multicast, data is diffused throughout the network
Disruption-tolerant networking (DTN) technologies
through the intermediate users who are also interested
enable mobile users to continue their communication
in the data content.
with surrounding devices even in the absence of a global
connection. Among different technologies developed We have implemented our interest-aware content dis-
for communication in mobile DTNs, routing algorithms tribution algorithm in The ONE [2] simulator for mobile
specifically handle data delivery from a stand-alone mo- DTNs. We ran several experiments using the reality-
bile source to its destination to mitigate the absence of mining database [1], which is one of the largest real-
an end-to-end connection. world mobile phone traces collected in academia. Our
Most ofthe routing protocols that have been proposed experimental results show that, compared to epidemic
for mobile DTNs address the data routing problem be- routing, our algorithm significantly increases the amount
tween a source and a specific destination [15, 20, 18, of data of interest received by the mobile users. This in-
12, 5]. However, in many real-world scenarios it is nec- creased coverage is achieved without distributing irrele-
essary to extend data transmission from individual re- vant data among those users.
cipients to a group of recipients. Proposed models for In Section 2 we motivate the necessity of an interest-

978-1-4244-4439-7/09/$25.00 ©2009 IEEE

aware content distribution protocol for mobile DTNs.
Section 3 briefly talks about related work in this area.
We discuss the details of our interest-aware routing al-
gorithm in Section 4. Section 5 explains our implemen- D
tation in the ONE simulator and our choice of mobility S -------
traces. In Section 6 we evaluate our simulation results in " , 1
detail. In Section 7 we review our future directions and
Section 8 concludes the paper.

2 Motivation - - - - - - wireless contact

- - - - _ , data transfer

In some application scenarios, a message needs to be Figure 1: Example of the interest-aware communication model. 1 - The
transmitted to every mobile node located in the DTN. source (S) estimates the potential recipients of data content (D) as belonging to
community Y . 2 - After calculating the similarity between the interests of the
For example, in emergencies, traffic congestion notifica- encountered users and data content D , the data is transferred to the most similar
tions, or severe weather alerts, all mobile nodes present user to b'fOUP Y which is node b. 3 Node b meets new mobile users . 4 - After
similarity calculations, two mobile nodes are selected as the interested recipients
in the area must be notified. In these scenarios, all mo- of the data content, nodes e and g .
bile users can receive the data through epidemic routing.
On the other hand, there are situations where not every- age and communication bandwidth. Also, the contact
one needs to receive a specific message and the mes- times between mobile users might be very short, espe-
sage or data content should be directed only to interested cially in highly mobile scenarios. Short contact times
users. For example, market news, sports events, sci- as well as limited bandwidth might disrupt the com-
entific articles, or advertisements about particular prod- plete transmission of the cached content. Our proposed
ucts, are of intent to a limited audience. In these ap- interest-aware algorithm, on the other hand, limits the
plications, an intelligent routing algorithm must be em- transmitted data to only the content of interest for mo-
ployed to identify the interested recipients of a specific bile users.
data content. An example scenario of an interest-aware communi-
The problem of multicasting has been studied in mo- cation model is presented in Figure I. In this scenario,
bile disruption-tolerant networks to handle communica- two different communities of users are identified as X
tion with groups of users [21]. This multicast architec- and Y. In this particular example, community Yare
ture is based on group membership model, where mobile users who are interested in the data content D. The al-
users register with a specific group to exchange multi- gorithm must recognize that the interested recipients of
cast messages with other group members. However, im- the data content D are users in group Y.
plementation of this model is infeasible in mobile DTNs
since there is no infrastructure to track group member- 3 Related work
In this paper, we propose an interest-aware communi- A major class of the routing algorithms for mobile
cation model that does not require users to obtain group DTNs is based on single-source, single-destination com-
memberships. Instead, our algorithm learns users' inter- munication model. These algorithms use some heuris-
ests from their previously downloaded and cached data tics to locate a specific recipient of a message through
content and properly assigns them to appropriate com- the mobile DTN. Message-ferrying [20], RPLM [18],
munities. Furthermore, our algorithm uses this inferred practical routing [12], and PROPHET [15] fall into this
information to efficiently provide each user with her fa- class. Our interest-aware routing algorithm, on the other
vorite content. hand, is designed to communicate with communities of
Our interest-aware communication model falls be- recipients.
tween unicast and epidemic multicast communication Communication with a group of recipients in mobile
models. Epidemic routing, assuming unlimited commu- DTNs has been discussed for multicast [21, 7]. This
nication resources, has been proved to yield the optimal multicast architecture requires tracking of the mobile
performance in terms of maximum coverage and min- nodes ' group memberships. The multicast group mem-
imum delay in mobile DTNs [19, 17]. This optimal berships are managed by special nodes such as ferries in
performance is achieved through the fact that all mo- the message-ferrying approach [7]. Designating special
bile nodes exchange all their cached data with each other nodes to handle group memberships, however, is not a
upon meeting. Therefore , epidemic routing is far from practical solution for mobile DTNs because nodes are
satisfactory because it creates a lot of unwanted traffic. joining and leaving the network .
Moreover, in reality mobile devices are limited in stor- A few studies have addressed the content distribution
problem in mobile DTNs. Leguay et al. [14] and 7DS's resentations to low-rank algebraic vectors of interests.
file synchronization application [16] use opportunistic Our model uses these interest-vectors to cluster users
contacts to distribute the data content among mobile into appropriate communities of interest. Because these
users in the mobile DTN. Both these architectures aim vectors are extracted from a group of documents in a
to maximize data dissemination epidemically without user's cache, they can be used to map a new document
considering mobile users' interests or social contexts. to a community of users who are interested in that docu-
Context-aware file sharing [8] application tries to share ment. This fact has been exploited in our interest-aware
files among users based on their interests. In this ap- routing algorithm to route documents to their interested
plication, users' interests are inferred from the exten- recipients.
sions of the files which are stored in their mobile de-
vices' cache. But files with different extensions could 4.1 Singular Value Decomposition
still contain similar content, the fact that is considered
in our interest-aware algorithm. LSA converts the problem of comparing and clus-
HiBOp [5], Profile-cast [10] and MobySpace [13] tering textual data to a problem of comparing algebraic
consider mobility behavior or previously visited geo- vectors in a multidimensional space using vector space
graphical locations by mobile users in their forwarding model (VSM) [4]. In the VSM, every unique term
decisions. They try to route a message closer to its des- (word) from the collection of the documents forms a
tination by forwarding the message to the relays with separate dimension and each document is represented
similar mobility pattern to the destination. These algo- by a vector spanning all these dimensions. The relation-
rithms limit the potential recipients to the mobile users ship between terms and documents is best expressed as
with some correlation in their roaming history. These a term-document matrix. Each element of this matrix is
models, therefore, are oblivious to the fact that users a numerical representation of the frequency of a term in
with uncorrelated mobility patterns might actually have the corresponding document.
some common interests. LSA employs singular value decomposition (SVD)
to reduce the rank of the term-document matrix in or-
Hui and Crowcroft [11] define different communities
of users by explicitly assigning their devices different der to eliminate insignificant, noisy words from the data
[4]. SVD produces a low-dimensional summary of the
labels. They show this user labeling can significantly re-
term-document matrix that exposes the underlying la-
duce the delivery cost, without trading off much against
tent concepts of the documents. SVD singularly breaks
the delivery ratio. Our interest-aware algorithm, how-
down the t x d matrix A into three matrices U, E, and
ever, automatically learns users' communities based on
users' interests which are latent in their cached data.
V, such that A == UE V T . U is a txt orthogonal matrix
whose column vectors are called left singular vectors of
A, V is a d x d orthogonal matrix whose column vec-
4 Interest-aware routing algorithm tors are called the right singular vectors of A, and E is
a t x d diagonal matrix having the singular values of A,
The first step in routing data content to mobile users ordered decreasingly along its diagonal. The rank r A of
based on their interests is to cluster them into communi- matrix A is equal to the number of its non-zero singular
ties of interests. User and document clustering has been values. By using the k largest singular values of A or k-
studied extensively in the context of web usage mining largest singular triplets of A k == UkEk VkT the original
[9]. Personalized information delivery, in this context, is term-document matrix A is approximated by A k . The
achieved by profiling people based on their web activi- first k columns of U, represented as Ui; form an orthog-
ties. For example, people are categorized based on the onal basis for this new approximated space. Orthogonal
hyperlinks they followed or the content ofread or down- columns of Ui; form the basis of a significantly reduced
loaded documents. One of the powerful methodologies dimensional feature space or concepts of the original
to discover groups of related people based on their re- matrix A. In other words, Ui; generated from stored
viewed documents is latent semantic analysis (LSA) [4]. documents in the user's cache, characterizes the user's
Using LSA technique, first, a low-dimensional topic- underlying interests. Therefore, we call the orthogonal
based representation ofweb documents is obtained. This vectors ofUkEk -1 interest-vectors and we refer to them
is followed by a construction of categories by clustering as Ik. The right multiplication by diagonal matrix Ek -1
such representations. differentially weighs the separate dimensions.
We have extended LSA to represent users' interests
in mobile DTNs based on documents they have reviewed 4.2 Interest-aware content distribution
or locally cached in their devices. In our model, we use
singular value decomposition (SVD) as one of the im- The fundamental concept behind the interest-aware
portant tools employed by LSA to summarize these rep- content distribution algorithm is based on matching the
~ 1 .r;) algorithm as a new module to the ONE's routing pack-

7J~ ~
age. The original implementation of the ONE simula-
tor was based on single-source, single-destination com-
I~ 1£
munication model and there was no notion of document
_ _ _ _3.;....~- categories. Therefore, we have extended the ONE sim-
correlation(D , Jt ) > ~ "-- --"'I ulator's source code to implement group-based commu-
cache i cachej nication and to create data content categories.
During each simulation round, we randomly gener-
ate a sub-population of mobile users as the target au-
Figure 2: Interest-aware content exchange 1 - Nodes i ,) discover each other. dience of a specific data content. We adjust this sub-
2 - Node i receives node)'s interest vectors. 3 - Node i relays document D to the population's interest-vectors to distinguish them as the
node ) ifnode)'s interests match with the document D.
target recipient's of that specific data. We, also, gen-
erate a finite number of communities, each of which
documents against the mobile node's interests. Based on is interested in some categories of data. The rest of
our algorithm, when two mobile users meet, the sender the population of mobile users is assigned some ran-
of the document D generates and transfers a copy of dom interest-vectors. In our simulation, in each random
the document to the other user if the document's con- timeslot, some mobile nodes are selected arbitrarily as
tent matches the other user's interests. In order to de- the senders of some data document. The document con-
termine this correspondence, document D should be tent is synthesized in the sender in order to meet one
mapped into the k-dimensional interest space of the re- of the communities' interests. One restriction that we
cipient. This mapping is represented as the inner prod- apply to choose the sender of a specific document is to
uct of the document and the recipient's interest vec- select it from the entire population excluding the target
tors; jj = DTUkL.k - 1 , where D is the vector of the recipients of that document. In our simulation scenar-
words in the document multiplied by the user's interest- ios, the document size is fixed and every mobile user
vectors. In fact, the elements of vector jj are inner has a limited cache size which is set to hold a maximum
products between vector D and columns of the matrix number of 5 million documents. The implementation of
UkL.k- I == h. Therefore , the normalized elements of the forwarding algorithm is based on the interest-aware
the vector jj represent the cosine similarities between communication model described in Section 4. Based on
the original document D and the user's interest-vectors. this algorithm, a document is transfered to the mobile
Document D is transferred to the other mobile user if users who have similar interests to the target recipients
there is enough correlation between the document's la- of that document.
tent concepts and the user's interest space. The steps The simulation evaluates the performance of the
of the interest-aware content exchange algorithm are interest-aware forwarding algorithm in distributing as
shown in the Figure 2. Node i sends the document D many relevant documents as possible among the users
to node j if the correlation between this document and who are interested in those documents' content. We
node j's interests is greater than some threshold value. compare the performance of our algorithm with the epi-
demic distribution of data content. We have chosen epi-
5 Simulation setup demic routing as our comparison ground for two rea-
sons. First, because , assuming unlimited communica-
We have evaluated our interest-aware content distri- tion resources such as storage and bandwidth, epidemic
bution protocol by simulations . The most important fac- routing is proven to produce maximum data distribution.
tor in a meaningful evaluation of the routing protocols The second reason is all other important routing algo-
by simulation is how well the assumptions about the for- rithms which are proposed for mobile DTNs don't ad-
warding algorithm and synthesized mobility traces re- dress data transmission to a group of recipients. In our
flect the real-world traits. In this section we explain our simulations , the movement of the mobile users is based
interest-aware protocol implementation in the ONE [2] on the reality-mining traces.
simulator and our choice of mobility traces for its per-
formance evaluation. 5.2 Reality-mining data traces

5 .1 Interest-aware protocol implemen- Previous studies show that the performance of mo-
tation in the ONE simulator bile DTN routing algorithms heavily depends on the
choice of the mobility model[3, 6]. DTN forwarding al-
The ONE simulator is a discrete event simulation en- gorithms demonstrate different performance results us-
vironment that is specifically designed for mobile DTN ing synthetic or real-world mobility traces. Therefore ,
scenarios. We have added our interest-aware forwarding in order to achieve a more realistic evaluation of our
interest-aware routing algorithm , we have decided to use 900 ,---- - ,---- - -,---- - -,---- - --,--- - ---,--- - ---,
the reality-mining database [I].
This dataset, which is a MySQL relational database ,
was collected by monitoring phone usage behavior of 70 ir- .--._._..,. . ,.. -.
800 ....-- .

the participants in order to model complex social sys-
tems. Among different information recorded in this
database , we used the Bluetooth devices encounter logs.
Based on our extracted statistics from these encounter 400
logs, the total number of Bluetooth devices recorded
in this study is 20,795 . Out of these 20,795 recorded 300
_ _ - I nterest ~aw are l

Bluetooth devices, 100 of them are Nokia 6600 mo- 1~ Epldemlc I

bile phones that belong to the participants in the reality-
mining experiment. Therefore, we extracted two dif- 50 100 150
Duration of contacts (sec)
200 250 300

ferent types of contact events from these traces. The

"participant-participant" contact events comprise the
contact events and their durations that occurred just Figure 3: Coverage of the commun ity of interest.

between the reality-mining participants. On the other

1200 ,-----~--~--~--~--~-------,
hand, the "Bluetooth-Bluetooth" contact events contain
both the encounters between the reality-mining partic-
ipants as well as contact events between participants ' 1000

and non-participants' devices. Then, we converted these
two extracted traces to a format accepted by the ONE 800

simulator's movement package . Considering both these

traces in evaluating our interest-aware routing algorithm

--.--.---......--....,...,.,.,.. .-.-- • - - - .- -

enables us to observe the role of data exchange with 500

external devices who are not participants in the reality- 400

mining study in dissemination of information among the 300

reality-mining community.
50 100 150 200 250 300
Duration of contads (sec)
6 Evaluation of results

In our simulations we are interested in data ex- Figure 4: Distribution of irrelevant documents among mobile users .
changes among mobile users participating in the real-
ity mining project or, in other words, our participant- ate number of 15 general categories , this number is not a
participant contact traces. But we have also stud- defining factor in the performance of our algorithm. All
ied the role of the encounters between participant and documents are randomly generated from some sources
non-participant devices using our extracted Bluetooth- which don't belong to the community of target recipi-
Bluetooth traces. ents. We specifically track the distribution of those doc-
In our simulations, we randomly select a commu- uments whose content is closer to the interests of our
nity of interest as a fixed percentage of the total reality- pre-specified target community.
mining participants , and we assign them similar interest- Contact traces of the reality-mining database have
vectors . We changed this size to vary from 10% to been recorded with the temporal resolution of 300 sec-
60% of the total number of the participants. Because onds. The analysis of the reality-mining traces reveals
we didn't see any significant difference in the final eval- that almost 44% of the observed contacts, that might
uations, the graphs represented here are related to the have lasted less than 300 seconds, are recorded with zero
community size of 40% of the total participants. This duration. So, in running the simulations for the original
subpopulation of mobile users is implicitly considered reality-mining data traces, there will not be any data ex-
as the target recipients. Furthermore, in the simulations change in 44% of the contact opportunities. Therefore,
we have specified a fixed number of general categories as well as running our simulations for the original traces,
for the documents and mobile users are randomly as- we have increased each contact duration by 30 seconds
signed to have some interest in each of these categories. at each round. This can be interpreted as increasing the
In reality, the number of categories of interests highly co-residence time ofmobile nodes in the same proximity
depends on the mobile users' demography. Although which is the case for networks with fewer movements or
our simulation results have been presented for a moder- mobile devices with longer radio ranges than those for
document exchange to only the documents with closer

:1 ' ' = " ,' .] content to the interests of the target recipients. There-
fore, interest-aware algorithm makes more efficient use

JOOIJ---- -- -- -------------_.
'50r ' • I of communication resources in terms of cache and wire-
less inter-contact times. Mobile devices have a lim-
ited cache size of 5 million documents. By prevent-
ing the distribution of irrelevant data, the documents in-
200 tended for the target recipients are better saved from be-
ing tossed out at intermediate nodes due to the cache
overflow. This fact has been further clarified in Figure
5. This figure shows that the number of dropped doc-
500:--- ----::
50- - --=--
- -=--
250 --::'
uments in mobile devices' cache using our protocol is
Duration of contacts (sec)
on average 16% lower than epidemic distribution. Fur-
thermore, since the exchanged documents in each wire-
Figure 5: Dropped document s in mobile device s due to cache overflow. less contact is limited to just the relevant documents, the
number of unfinished data transfers due to the termina-
tion of the connection decreases significantly, compared

I~L_. ~ ,' -1
to epidemic routing. Figure 6 shows the number of un-
necessary exchanges in each wireless contact opportu-
nity, on average, decreases by 35% using the interest-
aware algorithm.
j .--.--......-.-. 6 .2 Evaluation of the protocol using
Bluetooth-Bluetooth traces

50 In these simulations we used the contact events that

occurred between reality-mining participants and non-
00: --- ----::
50- - --=--
- -=--
Duration of contacts (sec)
250 --::'
300 participant Bluetooth devices or in another word the
Bluetooth-Bluetooth contact traces. We initially ex-
pected to achieve a better performance result in terms of
Figure 6: Aborted transfers due to insufficient contact times . the number of received relevant documents by the com-
munity of interest. This expectation was due to the fact
Bluetooth. that contact opportunities between participant and non-
participant devices should enhance the document distri-
6.1 Evaluation of the protocol using bution. But, after investigating the performance results
participant-participant traces we didn't observe much improvement in this regard. By
For these series of the simulations , we used collecting more statistics from the reality-mining origi-
participant-participant contact records between the nal database , we realized that out of the total number of
reality-mining participants. Figure 3 shows the num- 20,692 non-participant devices, 14,029 were just seen
ber of documents of interest that are received by the once in the entire trace log. In other words, 68% of the
mobile users in the target community after each sim- non-participant devices were never encountered again
ulation run. The abscissa represents contact durations by any of the reality-mining participants. Therefore,
that are increased in each run from zero to 300 seconds . these external contacts contribute very little to the data
As we see in the graphs, the number of distributed doc- distribution among reality-mining participants.
uments increases as the duration of contacts increases.
This is the natural outcome of providing more opportu- 7 Future Work
nities for mobile users to exchange documents. But, as
we can see, on average, our interest-aware content dis- We are planning to implement our interest-aware
tribution protocol distributes 30% more relevant docu- routing algorithm and release it for mobile devices. Us-
ments, with respect to the target community's interests . ing our algorithm in realizing the correlations among
This happens while interest-aware protocol distributes mobile users helps to understand users' social connec-
35% fewer irrelevant documents compared to epidemic tions from a conceptual level rather than just a spatial as-
routing, as represented in Figure 4. pect. Moreover, combining our algorithm with content
This is due to the fact that our algorithm limits the distribution applications [16] gives a better perspective
about the way data propagates among different commu- [8] M. Conti, F. Delmastro, and A. Passarella. Context-aware
nities of mobile users. Real-world implementation of file sharing for opportunistic networks. Mobile Adhoc
our algorithm enables to track the type of data and dif- and Sensor Systems, 2007. MASS 2007. IEEE Interna-
ferent paths it takes to distribute across different com- tonal Conference on, Oct. 2007.
[9] M. Grear, D. Mladenic, and M. Grobelnik. User profil-
munities. Therefore, it leads to a better understanding of
ing for the web. Computer Science Information Systems,
mobile users' inter-community interests. 3(2),2006.
[10] W. Hsu, D. Dutta, and A. Helmy. Profile-cast: behavior-
8 Conclusion aware mobile networking. SIGMOBILE Mob. Comput.
Commun. Rev., 12(1):52-54,2008.
Interest-aware content distribution protocol for mo- [11] P. Hui and 1. Crowcroft. How small labels create big im-
bile DTNs is capable of inferring mobile users' inter- provements. In PERCOMW '07: Proceedings ofthe Fifth
ests from their cached content. Our algorithm uses this IEEE International Conference on Pervasive Computing
inferred information to provide mobile users with data and Communications Workshops, pages 65-70, Washing-
content that matches their interests. Interest-aware con- ton, DC, USA, 2007. IEEE Computer Society.
[12] E. P. C. Jones, L. Li, and P. A. S. Ward. Practical routing
tent distribution algorithm uses singular value decom- in delay-tolerant networks. In WDTN '05: Proceedings
position (SVD) technique, as part of LSA algorithm, ofthe 2005 ACMSIGCOMM workshop on Delay-tolerant
to produce a low-dimensional summary of the docu- networking, pages 237-243, New York, NY, USA, 2005.
ments. Furthermore, this low-dimensional representa- ACM.
tion is used to cluster documents into appropriate cate- [13] J. Leguay, T. Friedman, and V. Conan. DTN routing in
gories. The users' categories of interest are represented a mobility pattern space. In WDTN '05: Proceedings of
by their interest-vectors. When two mobile nodes meet the 2005 ACM SIGCOMM workshop on Delay-tolerant
networking, pages 276-283, New York, NY, USA, 2005.
these interest-vectors are exchanged and used to calcu-
late the user's interest in the offered documents by the [14] 1. Leguay, A. Lindgren, 1. Scott, T. Friedman, and
other user. J. Crowcroft. Opportunistic content distribution in an ur-
We implemented our interest-aware protocol in the ban setting. In CHANTS '06: Proceedings of the 2006
ONE simulator and ran our simulations using the reality- SIGCOMM workshop on Challenged networks, pages
mining traces. Our simulation results show that our 205-212, New York, NY, USA, 2006. ACM.
protocol, comparing to the epidemic routing, distributes [15] A. Lindgren, A. Doria, and o. Schelen, Probabilistic
routing in intermittently connected networks. SIGMO-
30% more documents of interest among mobile users.
BILE Mob. Comput. Commun. Rev., 7(3):19-20,2003.
Furthermore, our interest-aware algorithm, on average, [16] A. Moghadam, S. Srinivasan, and H. Schulzrinne. 7DS -
has 35% less distribution of irrelevant data. a modular platform to develop mobile disruption-tolerant
applications. In Next Generation Mobile Applications,
References Services and Technologies, NGMAST 2008, Wales, UK.,
September 2008.
[1] MIT media lab reality mining. [17] S., K. Fall, and R. Patra. Routing in a delay tolerant network. In SIGCOMM '04: Proceedings of the 2004
[2] The ONE: The opportunistic network environment simu- Conference on Applications, technologies, architectures,
lator. dtn/theone, and protocols for computer communications, pages 145-
[3] M. Abdulla and R. Simon. The impact of the mobility 158, New York, NY, USA, 2004. ACM.
model on delay tolerant networking performance analy- [18] D. Thakore and S. Biswas. Routing with persistent link
sis. Simulation Symposium, 2007. ANSS '07. 40th An- modeling in intermittently connected wireless networks.
nual, pages 177 - 184, March 2007. IEEE Military Communications Conference, MILCOM
[4] M. W. Berry, S. T. Dumais, and G. W. O'Brien. Using 2005, pages 461--467 Vol. 1, Oct. 2005.
linear algebra for intelligent information retrieval. SIAM [19] A. Vahdat and D. Becker. Epidemic routing for par-
Rev., 37(4):573-595, 1995. tially connected ad hoc networks. In Technical Report
[5] C. Boldrini, M. Conti, 1. Jacopini, and A. Passarella. Hi- CS-200006, Duke university, April 2000.
bop: a history based routing protocol for opportunistic [20] W. Zhao, M. Ammar, and E. Zegura. A message ferrying
networks. IEEE International Symposium on a World of approach for data delivery in sparse mobile ad hoc net-
Wireless, Mobile and Multimedia Networks, 2007. WoW- works. In Mobilloc '04: Proceedings ofthe 5th ACM in-
MoM 2007, Helsinki, Finland, June 2007. ternational symposium on Mobile ad hoc networking and
[6] A. Chaintreau, P. Hui, C. Diot, R. Gass, and 1. Scott.
computing, pages 187-198, New York, NY, USA, 2004.
Impact of human mobility on opportunistic forwarding
algorithms. IEEE Transactions on Mobile Computing, [21] W. Zhao, M. Ammar, and E. Zegura. Multicasting in
6(6):606-620, 2007. delay tolerant networks: semantic models and routing
[7] Y. Chen, 1. Yang, W. Zhao, M. Ammar, and E. Zegura.
algorithms. In WDTN '05: Proceedings of the 2005
Multicasting in sparse MANETs using message ferrying.
ACM SIGCOMM workshop on Delay-tolerant network-
Wireless Communications and Networking Conference,
ing, pages 268-275, New York, NY, USA, 2005. ACM.
2006. WCNC 2006. Las Vegas, NV, 2:691---696,2006.