Milcom 2015 Track 2 - Networking Protocols and Performance

A Load Balanced Social-Tie Routing Strategy for DTNs Based on Queue Length Control Tuan Le, Mario Gerla Dept. of Computer Science, UCLA Los Angeles, USA {tuanle, gerla}@cs.ucla.edu

Abstract—Delay Tolerant Networks (DTNs) are sparse mobile ad-hoc networks in which there is typically no complete path between the source and destination. Although many routing algorithms for DTNs have been proposed, prior works generally focus on optimizing delivery ratio and cost by finding the relay node with the highest delivery probability to the destination. In social networks, the connections among nodes, which are established via social contacts, exhibit a fat-tailed distribution in which few nodes have many connections and the majority have very few. Since current heuristic-based routing algorithms bias toward connectivity, highly connected nodes have a high probability to be selected by other nodes as their next hop. As a result, the load distribution becomes significantly unbalanced, with very few nodes handling the majority of message forwardings. In this paper, we provide empirical experiments to show the natural load imbalance of existing routing algorithms. We then introduce Load Balanced Social-Tie Routing (LBR), a routing strategy in which messages are favorably forwarded to network nodes that have both a stronger social tie with the destination and a smaller or similar queue length. This queue length control strategy aims to reduce traffic at highly connected nodes, and allows nodes of similar degrees to trade packet forwardings with each other, thus spreading the traffic more evenly across the network. Through extensive simulation studies using a realworld San Francisco cab trace, we show that LBR can achieve a comparable or better delivery ratio and cost than existing algorithms. Meanwhile, LBR distributes the load more evenly with the top 10% of network nodes handling 23% of the forwardings, compared to 37% for Epidemic routing, 43% for PROPHET, and 47% for BubbleRap. Keywords—Delay Tolerant Networks; Social Networks; Routing Protocol; Load Balancing; Queue Length Control

I. I NTRODUCTION In Delay Tolerant Networks (DTNs) [1], mobile nodes contact each other opportunistically. Due to unpredictable node mobility, it is difficult to maintain persistent end-to-end connection between nodes. Thus, the store-carry-and-forward method is used for data transfers from source to destination. Node mobility is exploited to relay packets opportunistically upon contacting other nodes. A key challenge in DTN routing is to determine the appropriate relay selection strategy in order to minimize the number of message forwardings while maintaining a short message delivery time. Routing for DTNs has been widely studied in the past. To select the best relay node, a variety of network information, including dynamic network information (e.g., location information, traffic information, and encounter information), and static

978-1-5090-0073-9/15/$31.00 ©2015 IEEE

network knowledge (e.g., social relationships among nodes), is utilized. Compared with dynamic information, social ties and behaviors between nodes tend to be stable over time, and can be more reliably exploited to facilitate message transmission. Recently, many routing protocols have been proposed [2], [3], [4], [5], [6], which adopt social relationships among nodes to determine when and where to forward messages. These protocols are commonly known as socially-aware or sociallybased routing protocols. Although these works achieve a high delivery ratio, they often ignore the issue in distributing the network load evenly among network nodes. Achieving a balanced load distribution is critically important in DTNs in order to avoid network bottlenecks and a single point of failure, and to promote a fair contribution of nodes in forwarding and routing. In social networks, connections among nodes, which are established via social contacts, follow a fat-tailed distribution in which few nodes have many connections while the majority have very few (see Fig. 1). Social networks are an instance of a complex network [7], [8]. A random forwarding algorithm in a complex network has been shown to bias toward highly-connected nodes [9]. That is, a high-degree node has a high probability to receive a random-walk message from other nodes. As a result, highly-connected nodes have a high probability to be selected by other nodes as their next hop. This causes the load distribution to become highly unbalanced, with very few nodes carrying the majority of the traffic. Furthermore, if a forwarding algorithm is not random, but instead follows some heuristic that biases toward node connectivity, then the probability that a high-degree node receives messages will further increase. This further exacerbates the load imbalance among nodes. Existing works tend to forward messages via the relay node which has the highest delivery probability to the destination. In complex/social networks with fat-tailed node degree distribution, this strategy will guide the message toward highlyconnected nodes. Consequently, existing works trade an unbalanced network load distribution for a high delivery ratio. In this work, we propose a routing algorithm that aims to achieve a balanced load distribution, high delivery ratio, and low cost in terms of the number of forwardings. Our key idea is to use a queue length control mechanism, in which messages are only forwarded to network nodes that have a smaller or similar queue length. This strategy can reduce traffic at highly

383

Milcom 2015 Track 2 - Networking Protocols and Performance

Fig. 1. A social network graph with a fat-tailed degree distribution.

connected nodes, while allowing nodes to explore alternative (less congested) paths, thus spreading the traffic more evenly across the network. Furthermore, to achieve a high delivery ratio, we select relay nodes that have a higher social tie with the destination than the current node. Extensive simulation studies using a San Francisco cab trace show that our scheme distributes the load more evenly than existing schemes without compromising delivery ratio and cost. The paper makes the following contributions: • We identify the load balancing issue in existing DTN routing algorithms. • We propose a social-tie routing combined with a queue length control to achieve a better load distribution, while maintaining a high delivery ratio and a low cost. • We conduct extensive simulation studies to show the effectiveness of our scheme. The rest of the paper is organized as follows. Section II reviews the related work. Section III describes the routing strategy in detail. Section IV presents the experimental results. Section V concludes the paper. II. R ELATED W ORK Much work has been done regarding network architectures and algorithms for routing and forwarding in DTNs. Research on packet forwarding in DTNs originates from Epidemic routing [10], which floods the entire network. Spray and Wait [11] is another flooding scheme which restricts the number of copies. Recent studies develop relay selection techniques to approach the performance of Epidemic routing with a lower forwarding cost. Many schemes compute the delivery probability from the encounter node to the destination node before deciding whether to forward data. PROPHET [12] uses the past history of encounter events to predict the probability of future encounters. A node forwards data to its encounter only when the encounter’s delivery predictability is higher. MaxProp [13] is based on prioritizing both the schedule of packets transmitted to other peers and the schedule of packets to be dropped. These priorities are based on the path likelihoods to peers according to the historical information of node encounters. RAPID [14] considers DTN as a resource allocation problem. It optimizes a specific routing metric by translating it to per-packet utilities that determine at every

transfer opportunity if the marginal utility of replicating a packet justifies the resources used. CAR [15] and MV routing [16] consider the probability of staying at the same location (co-location) as the metric to find the relay. LeBrun et al. [17] use location information of nodes to forward data closer to the destination. Leguay et al. [18] observe that people that have similar mobility patterns are more likely to meet each other. Hence, they propose to forward data to nodes that have mobility patterns similar to the mobility pattern of the destination. Zhao et al. [19] take a different approach by utilizing a set of special nodes called message ferries (such as unmanned aerial vehicles or ground vehicles with short range radios) to help provide communication service for other nodes through the controlled non-random movements of the ferries. Since node mobility patterns are highly volatile and hard to control, attempts at exploiting stable social network structure for data forwarding have emerged. In [2], nodes are ranked using weighted social information. Messages are forwarded to the most popular nodes (highly-ranked nodes) given that popular nodes are more likely to meet other nodes in the network. The explicit friendships are used to build the social relationships based on their personal communications. HiBOp [20] labels each node with various contexts such as personal information, residence, and work. It then selects the next relay node according to nodes’ historical encounter records with the context of the packet destinations. SimBetTS [3] uses egocentric centrality and social similarity to forward messages toward the node with the highest centrality, to increase the possibility of finding the optimal carrier to the final destination. BubbleRap [4] combines the observed hierarchy of centrality and observed community structure with explicit labels to select the best forwarding nodes. The centrality value for each node is pre-computed using unlimited flooding. SMART [21] exploits a distributed community partitioning algorithm to divide a DTN network into smaller communities. For intra-community routing, SMART uses a utility function that combines both social similarity and social centrality for relay selection. For inter-community routing, SMART chooses nodes that move frequently across communities as relays. In [22], social features of each node are extracted, and the node that has more similar social features with the destination is selected as a relay node. In [23], relay selection is based on the local social map constructed using the knowledge of surrounding social structure. The delivery probability is computed over the entire path from source to destination. However, these works do not address the load balancing issue. They are designed to optimize for delivery ratio and cost by selecting the relay node with the highest delivery probability to the destination. This causes the majority of traffic to flow into highly-connected nodes, resulting in an unbalanced network load distribution. This paper combines information from both social-tie delivery probability and queue length to make the forwarding decision. The experimental results show a good balance between the load distribution and the delivery ratio and cost.

384

Milcom 2015 Track 2 - Networking Protocols and Performance

III. P ROTOCOL D ESIGN

B. Routing Strategy

In this section, we first describe the computation of the social tie, which implies the direct delivery probability between a pair of nodes. We then present a routing strategy based on the social tie and queue length control. A. Social Tie Computation In sociological terms, social tie describes an interpersonal connection by way of friendship or acquaintance. There are many tie strength indicators: frequency, intimacy/closeness, longevity, reciprocity, recency, multiple social context, and mutual confiding (trust) [3]. Among them, the most widely used heuristics in socially-aware networking applications are the recency and frequency of encounters [24]. Two nodes are said to have a strong tie if they have met frequently in the recent past. We compute the social tie between two nodes using the history of encounter events. How much each encounter event contributes to the social-tie value is determined by a weighing function F (x), where x is the time span from the encounter event to the current time. Assume that the system time is represented by an integer, and is based on n encounter events of node i. Then, the social-tie value of node i’s relationship with node j at the current time tbase , denoted by Ri (j), is computed as: Ri (j) =

n X

F (tbase − tjk )

We use a single-copy model in which, at any point in time, there is at most one copy of the message in the network. To achieve a high delivery ratio and low cost, we favor nodes that are good candidates to deliver the message successfully to their destination. In our routing strategy, a message carrier node i will forward the message to an encountered node j if and only if j has a higher social tie with the destination k than i. That is, the following condition must hold: Rj (k) > Ri (k)

In some cases, node i and its encounters may not have any social tie with the target destination k because, for example, they have never come into contact with k. Thus, the relay selection based on (2) can cause the message to get stuck in a node’s queue infinitely. To address this problem, we propose to forward a message to an encountered node that has a higher potential to deliver the message to any node. That is, X X Rj (x) > Ri (x) (3) x∈Nj

(1)

Ri (j) = F (10 − 1) + F (10 − 3) + F (10 − 5) = F (9) + F (7) + F (5) The weighing function F (x) essentially reflects the influence of the recency and frequency of encounter events. In order to give more weight to more recent encounter events, F (x) should be a monotonically non-increasing function. A function that satisfies this condition is F (x) = ( 21 )λx , where 0 ≤ λ ≤ 1. The control parameter λ allows a trade-off between recency and frequency in contributing to the socialtie value. As λ approaches 0, frequency contributes more than recency. On the other hand, as λ approaches 1, recency has higher weight than frequency. The social-tie value is solely determined by frequency when λ = 0, and by recency when λ = 1. In our experiments, the value of λ is carefully tuned based on the analysis of the network characteristic and is set to e−4 .

x∈Ni

where Ni and Nj are the set of nodes encountered by i and j, respectively. In summary, to optimize the delivery ratio and cost, node i only forwards a message intended for k to j if and only if any of the following conditions is met:   !  Rj (k) > Ri (k) P P   x∈N Rj (x) > x∈N Ri (x) ∧ (Ri (k) + Rj (k)) = 0

k=1

where F (x) is a weighing function, {tj1 , tj2 , · · · , tjn } are the encounter times when node i met node j, and tj1 < tj2 < · · · < tjn ≤ tbase . As an example, suppose node i met node j at times 1, 3, and 5, and that the current time (tbase ) is 10. Then, node i’s social-tie relationship with node j at tbase , denoted by Ri (j), is computed as:

(2)

j

(4)

i

However, this heuristic, just like any other existing heuristics, does not address the load balancing problem. In fact, it still biases toward high-degree nodes. To achieve load balancing, we use a queue length control mechanism such that traffic is temporarily diverted away from congested nodes (i.e., high-degree nodes). In this mechanism, nodes can only forward packets to nodes of similar or smaller queue length. That is, a congested node is allowed to forward packets to a less congested node, but not the other way around. The intuition behind this scheme is as follows: The queue length reflects a node’s connectivity. A highly connected node tends to receive lots of packets, and thus its queue length grows larger than others. By enforcing nodes to forward packets only to nodes of similar or smaller queue length, we can effectively divert traffic away from congested nodes, while allowing nodes to explore alternative paths. Over time, as packets flow out of congested nodes, their queue length becomes smaller, and the control mechanism will dynamically enable the traffic to flow into these nodes again. As we will show in Section IV, this queue length control strategy results in a more balanced load distribution without compromising delivery ratio and cost. With the queue length control, a node i will forward a message intended for k to j if any of the following conditions is met:

385

Milcom 2015 Track 2 - Networking Protocols and Performance

TABLE I. SIMULATION PARAMETERS  (Rj (k) > Ri (k)) ∧ (QLj ≤!QLi )    P P Rj (x) > Ri (x) ∧ (Ri (k) + Rj (k)) = 0  x∈Ni   x∈Nj ∧ (QLj ≤ QLi )

Parameter RxNoiseFigure TxPowerLevels TxPowerStart/TxPowerEnd m channelStartingFrequency TxGain/RxGain EnergyDetectionThreshold CcaModelThreshold RTSThreshold CWMin CWMax ShortEntryLimit LongEntryLimit SlotTime SIFS

(5)

QLi and QLj are the queue lengths of node i and j, respectively. IV. P ERFORMANCE E VALUATION In this section, we evaluate the performance of our proposed LBR scheme in a packet-level simulation, using a real-world mobility trace. We first describe the simulation setup, followed by the metrics used and the results. A. Simulation Setup We implement the proposed routing protocol using the NS3.19 network simulator. We adopt the IEEE 802.11g wireless channel model and the PHY/MAC parameters as listed in Table I. To obtain meaningful results, we use the real-life mobility trace of San Francisco’s taxi cabs [25]. This data set consists of GPS coordinates of 483 cabs, collected over a period of three consecutive weeks. For our studies, we select an NS-3 compatible trace file from downtown San Francisco (area dimensions: 5,700m x 6,600m) with 116 cabs, tracked over a period of one hour [26]. Vehicles advertise Hello messages every 100ms [27]. The broadcast range of each vehicle is fixed to 300m, which is typical in a vehicular ad hoc network (VANET) setting [28]. We evaluate the performance of LBR against the following algorithms: •

•

•

Epidemic routing [10] is a flooding-based multi-copy routing algorithm. It is optimal in terms of delivery ratio, but is very inefficient in terms of cost (the number of forwardings). Furthermore, Epidemic routing is expected to distribute the network load quite well as it does not apply any heuristic to guide the forwarding. Recall from previous sections that heuristics that select the relay with the highest delivery probability to the destination will bias toward highly-connected nodes, causing congestion and unbalanced load distribution. PROPHET [12] is a utility-based routing protocol that uses the past history of encounter events to forward data to nodes with higher delivery predictability to the destination. In our simulations, we use the same parameters as specified by the authors in [12]. That is, {Pinit , β, γ} = {0.75, 0.25, 0.98}. BubbleRap [4] is a community-based algorithm that routes data based on rankings calculated from the social centrality. A message is first bubbled up using the global ranking until it reaches a node in the same community as the destination. Then the local ranking is used to bubble up the message until the destination is reached or the message expires.

Value 7 1 12.5 dBm 2407 MHz 1.0 -74.5 dBm -77.5 dBm 0B 15 1023 7 7 20 µs 20 µs

In our experiments, each node sends a message to a random destination in the network after 1,000 seconds of the warmingup period. Each simulation is run for one hour. For statistical convergence, we repeat each simulation 20 times. B. Evaluation Metrics We use the following metrics for evaluation: • • •

Delivery ratio: the proportion of messages that have been delivered out of the total messages created. Cost: the total number of forwardings in the network. Load distribution: the distribution of the total number of forwardings across all network nodes.

C. Comparative Results Fig. 2a compares the delivery ratio among the schemes. As expected, Epidemic has the highest delivery ratio of around 65% after one hour of simulation. LBR and PROPHET deliver 48.7% and 45.2% of the messages, respectively. BubbleRap has a slightly worse performance with a delivery ratio of 41.5%. This is perhaps because BubbleRap is impacted by the weak community structure in the San Francisco cab trace. Recall that BubbleRap is a community-based algorithm, which makes forwarding decisions heavily based on the community structure of the network. In terms of the cost as shown in Fig. 2b, Epidemic routing and PROPHET require 151 and 1.15 times more forwardings than LBR. Although the cost of LBR is 1.2 times higher than BubbleRap, LBR has a better load distribution and delivery ratio than BubbleRap. Lastly, the load distribution is compared in Fig. 2c. LBR has the best load distribution with the top 10% of network nodes handling 23% of packet forwardings. This is significantly better than 37% for Epidemic routing, 43% for PROPHET, and 47% for BubbleRap. Note that Epidemic routing has a better load distribution than PROPHET and BubbleRap because it does not use any heuristic to guide its packet forwarding. BubbleRap has a worst load distribution because its forwardings are directed toward a few most popular nodes (highly-connected nodes) for the final direct packet delivery.

386

Milcom 2015 Track 2 - Networking Protocols and Performance

800

0.6

700 600

0.5

500

0.4 Cost

Delivery ratio

1 0.9

0.3

400 300

0.2 Epidemic LBR PROPHET BubbleRap

0.1 0 1000

1500

2000 2500 Duration (sec)

3000

3600

(a) Delivery ratio

200

Epidemic LBR PROPHET BubbleRap

100 0 1000

1500

2000 2500 Duration (sec)

(b) Cost

3000

3600

Percentage of total forwardings

0.7

0.8 0.7 0.6 0.5 0.4 0.3 Epidemic LBR PROPHET BubbleRap

0.2 0.1 0 0

0.2

0.4 0.6 Percentage of network nodes

0.8

1

(c) Load distribution

Fig. 2. Performance comparison of various routing strategies on the San Francisco cab trace.

V. C ONCLUSION In this paper, we proposed a routing protocol that makes forwarding decisions based on both the delivery probability of the encountered node in terms of the social tie, and its queue length. The effect of queue length control is to divert traffic away from congested nodes, and allow nodes to explore alternative, less congested paths to the final destination. Experimental results show that our scheme achieves better load balancing than existing schemes with the top 10% of network nodes handling 23% of the forwardings, compared to 37% for Epidemic routing, 43% for PROPHET, and 47% for BubbleRap. Furthermore, our scheme does not compromise the delivery ratio and cost; we are able to achieve a comparable or better delivery ratio and cost than existing schemes. R EFERENCES [1] K. Fall, “A delay-tolerant network architecture for challenged internets,” in Architectures and protocols for computer communications, 2003. [2] A. Mtibaa, M. May, C. Diot, and M. Ammar, “Peoplerank: Social opportunistic forwarding,” in INFOCOM, 2010 Proceedings IEEE. IEEE, 2010, pp. 1–5. [3] E. M. Daly and M. Haahr, “Social network analysis for information flow in disconnected delay-tolerant manets,” Mobile Computing, IEEE Transactions on, vol. 8, no. 5, pp. 606–621, 2009. [4] P. Hui, J. Crowcroft, and E. Yoneki, “Bubble rap: Social-based forwarding in delay-tolerant networks,” Mobile Computing, IEEE Transactions on, vol. 10, no. 11, pp. 1576–1589, 2011. [5] F. Li and J. Wu, “Localcom: a community-based epidemic forwarding scheme in disruption-tolerant networks,” in Sensor, Mesh and Ad Hoc Communications and Networks, 2009. SECON’09. 6th Annual IEEE Communications Society Conference on. IEEE, 2009, pp. 1–9. [6] E. Bulut and B. K. Szymanski, “Exploiting friendship relations for efficient routing in mobile social networks,” Parallel and Distributed Systems, IEEE Transactions on, vol. 23, no. 12, pp. 2254–2265, 2012. [7] R. Albert and A.-L. Barab´asi, “Statistical mechanics of complex networks,” Reviews of modern physics, vol. 74, no. 1, p. 47, 2002. [8] S. Boccaletti, V. Latora, Y. Moreno, M. Chavez, and D.-U. Hwang, “Complex networks: Structure and dynamics,” Physics reports, vol. 424, no. 4, pp. 175–308, 2006. [9] J. D. Noh and H. Rieger, “Random walks on complex networks,” Physical review letters, vol. 92, no. 11, p. 118701, 2004. [10] A. Vahdat, D. Becker et al., “Epidemic routing for partially connected ad hoc networks,” Technical Report CS-200006, Duke University, Tech. Rep., 2000. [11] T. Spyropoulos, K. Psounis, and C. S. Raghavendra, “Spray and wait: an efficient routing scheme for intermittently connected mobile networks,” in Proceedings of the 2005 ACM SIGCOMM workshop on Delaytolerant networking. ACM, 2005, pp. 252–259.

[12] A. Lindgren, A. Doria, and O. Schel´en, “Probabilistic routing in intermittently connected networks,” ACM SIGMOBILE mobile computing and communications review, vol. 7, no. 3, pp. 19–20, 2003. [13] J. Burgess, B. Gallagher, D. Jensen, and B. N. Levine, “Maxprop: Routing for vehicle-based disruption-tolerant networks.” in INFOCOM, vol. 6, 2006, pp. 1–11. [14] A. Balasubramanian, B. Levine, and A. Venkataramani, “Dtn routing as a resource allocation problem,” ACM SIGCOMM Computer Communication Review, vol. 37, no. 4, pp. 373–384, 2007. [15] M. Musolesi and C. Mascolo, “Car: context-aware adaptive routing for delay-tolerant mobile networks,” Mobile Computing, IEEE Transactions on, vol. 8, no. 2, pp. 246–260, 2009. [16] B. Burns, O. Brock, and B. Levine, “Mv routing and capacity building in disruption tolerant networks,” in INFOCOM 2005. 24th Annual Joint Conference of the IEEE Computer and Communications Societies. Proceedings IEEE, vol. 1, March 2005, pp. 398–408 vol. 1. [17] J. LeBrun, C.-N. Chuah, D. Ghosal, and M. Zhang, “Knowledge-based opportunistic forwarding in vehicular wireless ad hoc networks,” in Vehicular technology conference, 2005. VTC 2005-Spring. 2005 IEEE 61st, vol. 4. IEEE, 2005, pp. 2289–2293. [18] J. Leguay, T. Friedman, and V. Conan, “Evaluating mobility pattern space routing for dtns,” arXiv preprint cs/0511102, 2005. [19] W. Zhao, M. Ammar, and E. Zegura, “A message ferrying approach for data delivery in sparse mobile ad hoc networks,” in Proceedings of the 5th ACM international symposium on Mobile ad hoc networking and computing. ACM, 2004, pp. 187–198. [20] C. Boldrini, M. Conti, J. Jacopini, and A. Passarella, “Hibop: a history based routing protocol for opportunistic networks,” in World of Wireless, Mobile and Multimedia Networks, 2007. WoWMoM 2007. IEEE International Symposium on a. IEEE, 2007, pp. 1–12. [21] K. Zhu, W. Li, and X. Fu, “Smart: A social-and mobile-aware routing strategy for disruption-tolerant networks,” Vehicular Technology, IEEE Transactions on, vol. 63, no. 7, pp. 3423–3434, 2014. [22] J. Wu and Y. Wang, “Social feature-based multi-path routing in delay tolerant networks,” in INFOCOM, 2012 Proceedings IEEE. IEEE, 2012, pp. 1368–1376. [23] K. Chen and H. Shen, “Smart: Lightweight distributed social map based routing in delay tolerant networks,” in Network Protocols (ICNP), 2012 20th IEEE International Conference on. IEEE, 2012, pp. 1–10. [24] F. Xia, L. Liu, J. Li, J. Ma, and A. V. Vasilakos, “Socially aware networking: A survey,” 2013. [25] Cabspotting. [Online]. Available: http://cabspotting.org/ [26] J. Lakkakorpi, “ns-3 module for routing and congestion control studies in mobile opportunistic dtns,” in Performance Evaluation of Computer and Telecommunication Systems, 2013 International Symposium on. [27] M. van Eenennaam, W. K. Wolterink, G. Karagiannis, and G. Heijenk, “Exploring the solution space of beaconing in vanets,” in Vehicular Networking Conference (VNC), 2009 IEEE. IEEE, 2009, pp. 1–8. [28] S. Al-Sultan et al., “A comprehensive survey on vehicular ad hoc network,” Journal of network and computer applications, 2014.

387