This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
A Socially-based Routing Protocol For Delay Tolerant Networks Tamer Abdelkader∗ , Kshirasagar Naik† , Amiya Nayak‡ ,and Nishith Goel§ ∗ Department
of Electrical and Computer Engineering, University of Waterloo, Waterloo, Ontario N2M 3G1, Email:
[email protected] † Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Ontario N2M 3G1, Email:
[email protected] ‡ School of Information Technology and Engineering (SITE), University of Ottawa, Ottawa, Ontario K1N 6N5, Email:
[email protected] § Cistel Technology Inc., 40-30 Concourse Gatet, Ottawa, ON K2E 7V7, Email:
[email protected] Abstract—Networks in which nodes are intermittently connected, and have limited storage space and power, are termed Delay Tolerant Networks (DTN). To overcome these conditions, DTN routing protocols require nodes to store data packets for long periods of time until they contact with each other. In addition, they spread multiple copies of the same packet in the network to increase the probability of one of them reaching the destination. Long-term storage and multiple transmissions require large buffer space and non-restricted power availability which is hard to exist in DTN. In this paper, we study the routing problem in DTN with limited resources. We formulate a mathematical model for optimal routing, assuming the knowledge of present and future nodes contact and buffer space. After that, we analyze the previously developed heuristic protocols, and we propose a new protocol based on social relations between the nodes to avoid redundant copying of packets. Simulation results show that the proposed protocol significantly reduces energy consumption and provides better delivery ratio compared to other protocols.
and the inefficient use of the contact duration. Other routing protocols tend to restrict forwarding of data packets to selected nodes. Using some information collected about the network, they guide the packets to their destinations. This approach fails when the network topology is changing faster than the rate of information gathering. From the analysis of current DTN routing protocols, it can be concluded that an efficient protocol should efficiently spread limited copies of the same packet in the network to maximize the probability of reaching the destination, using the limited and uncertain information available. Therefore, our contributions in this paper are as follows: •
•
I. I NTRODUCTION Delay Tolerant Networks (DTN), are characterized by the lack of continuous end-to-end connections, and the limitedness of power sources and data storage space [1]. This type of networks has found its applications in many challenging environments such as providing delay-tolerant Internet services to suburban and rural areas [2], and to vehicles [3]. In addition, DTN have its promising applications in monitoring and tracking wildlife and whales in oceans [4], environmental monitoring such as lake water quality monitoring [5], and many others. Traditional routing protocols for wired and wireless networks failed to work in this environment because they assumed the existence of continuous end-to-end connection between sources and destinations. Routing protocols developed for DTN adapt themselves to this challenging environment by probabilistically sending multiple copies of data packets so that one of them may reach the destination. Nodes receiving the packets store them until they meet other nodes or meet their destinations. Simple DTN routing protocols blindly send data packets to the nodes they meet without having a selection criterion. They range from the full network flooding to the limited flooding. This approach has its drawbacks, such as burdening the buffer
•
•
A mathematical formulation of optimal routing is developed, assuming the availability of present and future nodes contact and buffer information. A heuristic routing protocol is proposed. The protocol exploits the social relationships between network nodes to increase the chance of reaching the destination, without flooding the network with too many copies . Using simulations, we compare the proposed protocol with a full flooding protocol (Epidemic), a limited flooding protocol (Binary Spray-And-Wait (SnW)) and a guided routing protocol (PROPHET) in addition to the optimal protocol in terms of their delivery ratio, number of transmissions, and average packet delay. The comparison is conducted under different conditions: the number of nodes in the network, their buffer capacities, the traffic load (number of packets generated), and their time-to-live (TTL) values. Results show that our proposed protocol efficiently spread packets over the network achieving higher delivery ratio with minimal energy consumption than other protocols.
The rest of the paper is organized as follows. In section II, we present the modeling of optimal routing as an optimization problem. The proposed heuristic protocol is presented in section III. The performance comparisons and simulation results are discussed in Section IV. A brief overview of related work is presented in section V. Finally, conclusions and future work are discussed in section VI.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
TABLE I S YMBOLS USED IN THE O PTIMAL ROUTING Symbol N M xc,m Bi bi C tc Dc dm Lm
Description Number of nodes in the network Number of messages in the network A variable indicating if message m is transferred during contact c, 1=yes, 0=No The buffer capacity of node i The buffer size at the time of optimization The set of contact objects The starting time of contact c The duration of contact c The transmission time for message m The lifetime (TTL) of message m
II. O PTIMAL ROUTING FOR DTN Our objective is to find a route for each message, if it exists, which minimizes the number of hops to destination. A route is an ordered set of contacts from source to destination. Each contact has four attributes: Sender, Receiver, Time, and Duration, where • • • •
Sender: The node which sends the messages. Receiver: The node which receives the messages. Time: The time at which the two nodes meet. Duration: The period length in which the two contacting nodes are able to transfer messages.
Each node buffer has two attributes: Capacity and size, where • •
Capacity: The maximum number of messages the buffer can carry. Size: The actual number of messages at the time of optimization.
•
•
• •
•
Buffer: The messages transferred during each contact should not exceed the receiver buffer capacity. Flow Conservation: The number of contacts in which each message is sent should be equal to the number of contacts in which that message is received, except for its source and destination of that message. Source and Destination: There should be only one contact in which each message is sent by its source, and one contact in which each message is received by its destination. Contacts order: Contacts in the same route should have increasing order of their starting times. Contact duration: The sum of all message transfer times during a contact should be less than the duration of that contact. Message Lifetime: The starting time of the last contact in the route of a message should be less than the message lifetime (Time-To-Live or TTL).
The symbols used in the formulation are described in table I. The problem formulation is shown next. A detailed explanation of the modeling can be found in [6].
m∈M
c∈C
xc,m
m∈M c1∈C xc1,m − Bc.r − bc.r,0 , if c1.r =
m∈M
c2∈C
xc2,m ≤
c2.s,
tc1 , tc2 < tc c1∈C xc1,m − c2∈C xc2,m = 0 ∀n ∈ N , ∀m ∈ M , if n = c1.r = c2.s x = 1, if c.s is the source c,m c∈C x = 1, if c.r is the destination c,m c∈C tc1 xc1,m + (H − tc2 )xc2,m < H, if c1.r = c2.s, ∀c1, c2 ∈ C, ∀m ∈ M H is a large number where H > tc1 , H > tc2 , m∈M dm xc,m ≤ Dc , ∀c ∈ C
Flow Conservation Source and Destination Contacts Order
Contacts Duration Message Lifetime
tc xc,m < Lm , ∀c ∈ C, ∀m ∈ M xc , xc1,m , xc2,m ∈ {0, 1}, tc1 , tc2 , dm , Dc , Lm ≥ 0
III. S OCIALLY-BASED ROUTING F OR D ELAY-T OLERANT N ETWORKS (SBR-DTN) Our proposed protocol can be explained as follows: • Each packet generated is assigned a unique ID that is associated with it and all its copies till they are deleted or they reach the destination. The list of all the packets IDs in a node’s buffer is called the summary vector. • Each node has a degree of connectivity to every other node that is strengthened by their frequent meetings. • When two nodes meet, they update their degree of connectivity, Γa,b , using the following equation: Γa,b = Γa,bold γ k + (1 − Γa,bold γ k )α
The problem solution should satisfy a set of constraints: •
Minimize subject to Buffer constraint
•
•
•
•
•
where Γa,b is the degree of connectivity between nodes a and b, α ∈ [0, 1] is the updating factor, γ ∈ [0, 1] is the aging constant, and k is the number of time units that have elapsed since the last time they met. The two nodes then exchange their summary vectors. All data packets that are stored in one node and not in the other are ordered on a first come first serve (FCFS) basis to be transferred to the other node. Packet transfer starts until the contact duration ends. The number of packets transfered is reversely proportional to the degree of connectivity. The idea is that the more the connectivity between two nodes, the more they are expected to meet, the less is the benefit that they carry the same list of packets, and vice versa. After a packet is transfered, it may be deleted from the sender node with a probability equal to the degree of connectivity. The number of data packets transfered is also limited, using the Binary SPRAY-and-WAIT (SnW) mechanism, so that a node keeps half the number of copies and assigns the other half to the receiving node. When the expiry time of a data packet approaches, the carrying node switches to partial flooding of the network
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
with limited copies of the packet. This gives a last chance to the dying packet to catch the destination before it is deleted. The pseudo-code for the proposed protocol is presented in Procedure 1. Procedure 1 OnContact Input: node: n1,n2;ContactDuration 1: Excchange Summary Vector(n1,n2) 2: Age = CurrentT ime − LastM eetingT ime(n1, n2) 3: Γa,b = Γa,bold γ Age + (1 − Γa,bold γ Age )α 4: if ContactDuration > 0 then 5: if pkt=GetPacket(n1) then 6: if NotExpired(pkt) and NotReceivedBefore(pkt,n2) then 7: if IsDestination(pkt,n2) then 8: ConsumePacket(pkt,n2) 9: else if ExpiryT ime < ExpiryT hreshold then 10: N rCopies=GetLastChance(pkt,n1) 11: if N rCopies > 1 then 12: StorePacket(pkt,n2) 13: SetLastChance(pkt,n1,N rCopies − 1) 14: SetLastChance(pkt,n2,1) 15: end if 16: else 17: U = U nif ormRandomN umber 18: N rCopies=GetNrOfCopies(pkt,n1) 19: if U > Γa,b and N rCopies > 1 then 20: StorePacket(pkt,n2) 21: SetNrOfCopies(pkt,n1,N rCopies/2) 22: SetNrOfCopies(pkt,n2,N rCopies/2) 23: U 2 = U nif ormRandomN umber 24: if U 2 < Γa,b then 25: DeletePacket(pkt,n1) 26: end if 27: end if 28: end if 29: ContactDuration = ContactDuration − 1 30: end if 31: end if 32: end if
IV. P ERFORMANCE C OMPARISON AND S IMULATION R ESULTS We consider a sparse mobile network where nodes are connected to each other at discrete time intervals via wireless links. Nodes communicate when they get into the communication range of each other. In such event, they are said to be in contact. The inter-contact time, that is the interval between two contacts of the same pair of nodes is modeled using the power law distribution [7]. The cumulative distribution function (CDF) of the power law distribution is as follows −θ x (1) P(X ≥ x) = xmin
where x is the inter-contact time, xmin is the minimum intercontact time and θ is the parameter that characterizes the power law, θ > 0. In our experiments, we use θ = 0.9 as mentioned in [7]. To avoid partitioning, each node is made to contact with at least one node. The maximum number of nodes to contact with is drawn from a uniform distribution of up to one fifth of the total number of nodes in the network. Packets are created at every node using Poisson distribution, and assigned a random destination. When generated, each packet is associated with a unique identifier, time of creation and a time to live (T T L). Table II shows the values used for the network parameters. TABLE II N ETWORK A ND P ROTOCOLS PARAMETERS Protocol ALL
PROPHET SnW, SBR-DTN SBR-DTN
Parameter Simulation time Minimum packet arrival time Minimum inter-contact time Initialization constant Transitivity constant Aging constant,γ Initial number of copies Updating factor Aging constant,γ Last-Chance number of copies Expiry threshold
Value 12 hours 1 minute 4 hour 0.75 0.25 0.98 4 0.45 0.98 10 0.1 ∗ T T L
Each node records all the packets it receives to avoid receiving two copies of the same packet. We study the effect of varying the buffer capacity (B), the packets TTL, the traffic load (T L) and the number of nodes in the network (N ) on the performance of the protocols: Epidemic, PROPHET, Binary SPRAY-AND-WAIT (SnW), the proposed protocol (SBRDTN) and the optimal protocol. The performance measures considered are: • • •
Delivery ratio: The number of delivered packets to the number of packets generated. Number of transmissions: The number of packet copies to the number of delivered packets. The average packet delay: The sum of delivered packets delays to the number of delivered packets.
A. Impact of Varying the Buffer Capacity (B) As shown in Figure 1, increasing buffer capacity increases both delivery ratio and average packet delay. This is justified by the reduction of dropped packets due to buffer overflow and, therefore, were able to reach destinations with longer delay. Changing the buffer capacity does not significantly affect the network overhead. Our proposed protocol, SBRDTN, achieves almost the same delivery ratio, with much less number of transmissions than the other heuristic protocols. The average packet delay of the optimal protocol is greater than the heuristic ones, because its main goal is to find valid routes. This proves that by tolerating more delay, we can increase delivery ratio and reduce energy consumption.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
B. Impact of Varying the Packet Time-To-Live (T T L) It can be noticed from Figure 2 that giving more lifetime to the packets increases both the delivery ratio and the average delay, because of the increased chance to reach destination during the longer life. This increase in delivery ratio is small in the heuristic protocols because the effect of dropping packets due to buffer overflow is more significant than giving longer lifetime to the packets. SBR-DTN maintains the lowest number of transmissions among all heuristic protocols for all TTL values. C. Impact of Varying the Traffic Load (T L) As shown in Figure 3, increasing the traffic load (number of packets generated in the network), causes buffers to overflow and stored packets to be deleted, which decreases the delivery ratio. Packets delivered are those transfered quickly between the nodes. Therefore, we find that the average packet delay decreases with increasing the traffic load. D. Impact of Varying the Number of Nodes (N ) Increasing the number of nodes increases their contact frequency and the total network buffer space, as shown in Figure 4. This leads to increasing both the delivery ratio and the number of transmissions. More buffer space increases the packets waiting time, and more contact frequency, decreases the packets waiting time in buffer. V. DTN ROUTING P ROTOCOLS Routing protocols are classified according to the amount and type of information used to take the routing decision [8]. Blind routing protocols aim at fast spreading of packets in the network. They do not use any information to select next nodes. They vary according to their spreading mechanism and amount. Epidemic [9] was historically the first routing protocol that belongs to this class. In Epidemic, each node spreads all the packets it has to all the other nodes it meets, until the packet expires or a destination acknowledgment is received. Because of the limited storage space and contact durations, the protocol performance drops significantly with the high traffic rates. To overcome this problem, other routing protocols limit the flooding of packets to a certain number of copies or hops. Spray-and-Wait (SnW) [10] protocol limits the number of copies by associating with each copy the number of extra copies to spread. When no more spreading is allowed, the carrying node keeps the packet until it either meets the destination or the packet is deleted due to buffer overflow or lifetime expiry. A binary version of SnW permits each node to use half the number of transfers allowed for the packet and the other half is left for the receiving node. Guided routing protocols use the available network information to guide packets to their destinations. These protocols assign weights to nodes using information such as expected contact times, location and mobility patterns of nodes, number of packet copies. This information can be collected using special infrastructure [4], [11] or by exchanging data between
mobile nodes [12], [13]. The PROPHET protocol [12] estimates a node metric by tracing the number of meetings between nodes. When two nodes meet, they increase their link weight towards each other and towards the nodes met by the other node. Guided routing protocols outperform blind protocols in the delivery ratio, but increases the average packet delay. VI. C ONCLUSION Delay Tolerant Networks (DTN), lack end-to-end connections between data sources and destinations. This require the intermediate nodes to store data packets for long periods of time which violates one of the basic assumptions of traditional routing protocols and triggers the development of new ones. In this paper, we provided a mathematical model for optimal routing in DTN under different constraints. In addition, we presented a heuristic routing protocol that utilizes the social relations between nodes to reduce redundant copying of packets. Simulation results show that the proposed protocol significantly reduces number of transmissions leading to a considerable saving in energy consumption, while keeping same or higher delivery ratio. The study also shows that increasing the delivery ratio causes an increase in the average packet delay which is acceptable in a delay tolerant networks. R EFERENCES [1] “Delay tolerant networking research group.” [Online]. Available: http://www.dtnrg.org [2] A. Pentland, R. Fletcher, and A. Hasson, “Daknet: rethinking connectivity in developing nations,” vol. 37, no. 1, 2004, pp. 78–83. [3] J. Ott and D. Kutscher, “Drive-thru internet: Ieee 802.11b for ”automobile” users,” vol. 1, 2004, p. 373. [4] T. Small and Z. J. Haas, “The shared wireless infostation model: a new ad hoc networking paradigm (or where there is a whale, there is a way),” in ACM MobiHoc, June 2003, pp. 233–244. [5] “Sensor networking with delay tolerance (sendt).” [Online]. Available: http://down.dsg.cs.tcd.ie/sendt/ [6] T. Abdelkader, K. Naik, and A. Nayak, “Choosing the objective of optimal routing protocols in delay tolerant networks,” in The 6th IEEE International Computer Engineering Conference, Cairo, Egypt, December 2010. [7] A. Chaintreau, P. Hui, J. Crowcroft, C. Diot, R. Gass, and J. Scott, “Impact of human mobility on opportunistic forwarding algorithms,” IEEE Transactions on Mobile Computing, vol. 6, no. 6, pp. 606–620, June 2007. [8] Z. Zhang, “Routing in intermittently connected mobile ad hoc networks and delay tolerant networks: overview and challenges,” Communications Surveys & Tutorials, IEEE, vol. 8, no. 1, pp. 24–37, March 2007. [9] A. Vahdat and D. Becker, “Epidemic routing for partially connected ad hoc networks,” 2000. [10] T. Spyropoulos, K. Psounis, and C. S. Raghavendra, “Efficient routing in intermittently connected mobile networks: the multiple-copy case,” IEEE/ACM Trans. Netw., vol. 16, no. 1, pp. 77–90, 2008. [11] R. C. Shah, S. Roy, S. Jain, and W. Brunette, “Data mules: Modeling a three-tier architecture for sparse sensor networks,” in IPSN 2008: Proceedings of the First IEEE Workshop on Sensor Network Protocols and Applications. IEEE, 2003, pp. 30–41. [12] A. Lindgren, A. Doria, and O. Schelen, “Probabilistic routing in intermittently connected networks,” SIGMOBILE Mobile Computing and Communication Review, vol. 7, no. 3, pp. 19–20, July 2003. [13] T. Abdelkader, K. Naik, and A. Nayak, “An eco-friendly routing protocol for delay tolerant networks,” in The 6th IEEE International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob 2010), Niagara Falls, Canada, October 2010.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
(a) Delivery Ratio
(a) Delivery Ratio
(b) Overhead Ratio
(b) Overhead Ratio
(c) Average Packet Delay Fig. 1.
(c) Average Packet Delay
Performance of the protocols, under different buffer capacities. Fig. 3.
Performance of the protocols, under different traffic loads.
(a) Delivery Ratio (a) Delivery Ratio
(b) Overhead Ratio (b) Overhead Ratio
(c) Average Packet Delay (c) Average Packet Delay Fig. 2.
Performance of the protocols, under different packet TTL values.
Fig. 4. Performance of the protocols, under different Network loads (number of nodes).
978-1-4244-5637-6/10/$26.00 ©2010 IEEE