compare with three protocols:
and Epidemic [3]. We make because Epidemic doesn’t con We set the area concept and l and reply action in Epidemic.
Then, We will show some explain how the equations can (1) Query-Reply Success R message / number of query me reply message means all the delivered from Replier to Que the same data repeatedly. Th means how many numbers the
(2) Query Delivery Ratio number of query message. T means the all query message source node.
(3) Overhead = (#relay mes
#query delivered)/(#success rep The messages which are redun Success reply message means Querier and the same messa double count. Query deliver delivered to the Replier, and don’t include.
(4) Latency = Time(Rec Time(sent first query message) sending the first query messag receiving the first reply messa the the medium number of the r
A. Simulation setup
We use the ONE(Opport simulator) [20] and the map of validate our approach. All nod
rithm
ery dataset Q_set, Reply dataset uery weight q, Reply weight r,
ize(Q_set) + size(R_set)) !=
etSet()
the simulation results, and we LCS (we proposed), Locus [11]
e some variation in Epidemic, nsider the query-reply situation.
et nodes having query scenario validation parameters. And we be calculated.
Ratio = number of success reply essage. The number of success e reply messages successfully erier. Note that we won’t count
he number of query message query message are generated.
= number of query deliverd / The number of query delivered es which delivered to the data sage - #success reply message - ply message + #query deliverd).
ndantly transferred are overhead.
s the first message reply to the ages replied afterword do not
red means the first message as above, the same messages ceived first reply message) – ). We calculate from the time of ge by the Querier to the time of age by the Replier. And we use
results for the analysis.
tunistic Network Environment f Taipei-101 surrounding area to des are pedestrian, and we set a
random destination point to them. When destination, then we give them the other de nodes move to the position by shortest path a shows the simulation settings.
Table 1: Simulation Settings Area
Interval of message creation Data message
Query message (offline-node only) Buffer size
Velocity
Time-To-Store(TTS) B. Simulation result
In Figure 3, we can see our approach a similar performance in query-reply success r are using the region concept to centralize th as the number of nodes increase, Locus will because our region concept is based on local many nodes in this area, it will have many query, reply) in every node. Although nod source, the transmission rate is fixed, intermittently connected, they can’t send all some messages might be ignored, and the p well. And we want to see the important par we modify it in two points. (1) We use ep replace the data replication strategy, and (2) routing to replace the query replication. The (1) type got worse performance. Because t couldn’t be centralized to the inside area, th spread to the whole network. It is difficult data message in the network.
On the other way, in order to compare fa scenario with only Offline-node in our appro success ratio is worse than Locus, but t latency are better than Locus.
Figure 3: Query-Reply Success Ratio (Nod
they reached the perform well. It is l area. If there are y messages (data, des have rich data and nodes are the messages. So, performance is not rt of our approach, pidemic routing to ) we use epidemic e result shows the the data messages he messages would t to query unique airly, we present a oach. Although the the overhead and
e Density)
Figure 4 is overhead, and w Although the query success rat but its overhead is larger tha centralizes the same data in the to this area, it can use a few quickly. But, if Querier is far replicate more message in del higher. And LCS is a wide ran all the messages in the area, th to spend lower cost to get the d
Figure 4: Overh Figure 5 is Query Latency approach LCS is much lowe Because all the data will be sp Inside Area or Border Area, Although Locus also uses the query the data which are not n time to query data.
Figure 5: Latenc Figure 6 shows the query-can see that if low TTS param has bad performance. It is bec time to live, they don’t have m the success ratio doesn’t go we is worse than LCS, it is beca centralize the data in its area, b The data is dropped quickly, data.
we can see the result of Locus.
tio of Locus is higher than LCS, an LCS. It is because Locus e same area. If Querier is close query replicate to look up data r away from this area, it has to livery. So the overhead will be nge of the local area, it spreads hus, LCS has higher opportunity data.
head (Node Density)
y. We can see the result of our er than Locus and Epidemic.
pread in the area, if node in the it will quickly get the data.
area concept, if nodes want to near the area, it will take much
cy (Node Density)
-reply success ratio results. We meters are used, every approach cause messages just have short much time to delivery such that ell. In the low TTS case, Locus ause they will try their best to but the TTS of data isn’t enough.
so nodes are difficult to query
Figure 6: Query-Reply Success Ratio ( Figure 7 shows the overhead results. As t the overhead of all the approaches will incr messages have longer time to stay in the n messages are successfully delivered to the message which are not yet sent to the Queri network. Then, these messages will increase
Figure 7: Overhead (TTS) Figure 8 shows the latency results. As th latency of all the approaches will increase. I are many messages in every node, and in th The old message will be sent first. So the ne to be sent, then, it could take much time to fo
Figure 8: Latency (TTS) V. CONCLUSIONS In this paper, we proposed a location-bas approach. We use four strategies to achie They are Data replication, Query replication Data synchronize and update strategies. Th three-tier area concept, and every strategy will have different rule to replicate message
(TTS)
the TTS grows up, ease. It is because network. If reply Querier, the other er still exist in the
the overhead.
he TTS grows, the It is because there he sending queue.
ew one is difficult orward.
sed content search eve our objective.
n, Data reply, and hen we proposed a y in different area es. We divided all
nodes into two types of statues If Offline-node wants to search other Offline-node, it can ask O in the sending message period selection algorithm to decide w Finally, we evaluate the results we use some parameters to v LCS is worse than Locus i however, in Overhead and Lat Locus.
REFER [1] V. Cerf, S. Burleigh, A. Hooke, L and H. Weiss, “Delay-Tolerant Netw 2007, [Online]. Available: http://www.
[2] K. Fall, “A Delay-Tolerant N Internets”, In Proc. SIGCOMM, Augus [3] A. Vahdat and D. Becker, “Epide hoc networks,” Tech. Rep. CS-2000-06 [4] T. Spyropoulos, K. Psounis, C.S efficient routing scheme for intermitte USC, USA, In Proc. SIGCOMM, Augu [5] T. Spyropoulos, K. Psounis, C Efficient Mobility-Assisted Routing Mobility”, USC, EECS, USA, In Proc.
[6] A.Lindgren, A.Doria, and O.
Intermittently Connected Networks”, vol. 7-3, July 2003.
[7] D.K. Cho, C.W. Chang, M.H.
monitoring in the battlefield”, CS, November 2008.
[8] C. Caini, P. Cornice, R. Firrincieli Approach to Satellite Communicati Bologna, Italy, In Proc. IEEE Journal vol. 26-5, June 2008.
[9] Y. Sasaki, Y. Shibata, “Distributed Based Mobile Communication Envi November 2010.
[10] P. Jinag, J. Bigham, E. Bodanes Emergency Communications with DT London, UK, In Proc. WCNC, March 2 [11] Nathanael Thompson, Riccardo C Location-based Data Overlay for University of Illinois Urbana-Cham September 2010.
[12] P. Yang, M.C. Chuah, “Perf information retrieval schemes for DTN Proc. MILCOM, November 2008.
[13] Cong Liu, Jie Wu, Xin Guan and Hybrid Delay Tolerant Networks”, In P [14] Mikko Pitk¨anen, Teemu K¨arkk¨a
“Searching for Content in Mobile DTN Finland, In Proc. PERCOM, March 20 [15] S.C. Nelson, M. Bakht, R. Kravet CS, University of Illinois at Urbana-C April 2009.
[16] E.C.R de Oliveria, C.V.N. de Alb Protocol Based on Neighborhood C Fluminense, Brasil, In Proc. SAC, Mar [17] G. Sollazzo, M. Musolesi, C.
content-based dissemination system University College London, UK, In Pro [18] C.E. Palazzi, A. Bujari, “A D Mobile-to-Mobile File Sharing”, Unive Proc. IFIP, October 2010.
[19] T. Vincenty, “Direct and Inverse with application of nested equations”, [20] Ari Keränen, Jörg Ott, Teemu DTN Protocol Evaluation”, In Proc. Si
, Online-node and Offline-node.
h some information, besides the Online-node for searching. And , we proposed a message queue which message will be sent first.
s with Epidemic and Locus. And verify our approach. Although in Query-Reply success ratio, tency, LCS is much better than
RENCES
. Torgerson, R. Durst, K. Scott, K. Fall working Architecture”, RFC4838, April
.ietf.org/rfc/rfc4838.txt
Network Architecture for Challenged st 2003.
emic routing for partially-connected ad 6, Duke University, July 2000.
S. Raqhavendra, “Spray and wait: an ently connected mobile networks”, EE, ust 2005.
C.S. Raqhavendra, “Spray and Focus:
g for Heterogeneous and Correlated PERCOM, March 2007.
Schel’en, “Probabilistic Routing in LUT, Sweden, In Proc. SIGMOBILE, Tsai, M. Gerla, “Networked medical UCLA, USA, In Proc. MILCOM, i, D. Lacamera, S. Tamagnini, “A DTN ions”, DEIS/ ARCES, University of
on Selected Areas in Communications, d Disaster Information System in DTN ironment”, Japan, In Proc. BWCCA, se, “Adaptive Service Provisioning for TN”, EECS, Queen Mary University of
2011.
Crepaldi and Robin Kravets, “Locus: A Disruption-tolerant Networks”, CS, mpaign, USA, In Proc. CHANTS, formance evaluations of data-centric Ns”, CSE, Lehigh University, USA, In d Li Chen, “Cooperative File Sharing in
Proc. ICDCSW, June 2011.
ainen, Janico Greifenberg, and J¨org Ott, Ns”, Helsinki University of Technology,
09.
s, “Encounter–Based Routing in DTNs”, Champaign, USA, In Proc. INFOCOM, buquerque, “NECTAR: A DTN Routing ontact History”, Universidade Federal rch 2009
Mascolo, “TACO-DTN: a time-aware for delay tolerant networks”, CS, oc. MobiOpp, June 2007.
Delay/Disruption Tolerant Solution for ersity of Padova, Italy, Wireless Day, In
Solutions of Geodesics on the Ellipsoid Survey Review, Vol.23, 1975.
Kärkkäinen, “The ONE Simulator for muTools, March 2009.