Topology aware Content Centric Networking for Mobile Ad Hoc ...

TOP-CCN: Topology aware Content Centric Networking for Mobile Ad Hoc Networks Jaebeom Kim, Daewook Shin, and Young-Bae Ko Department of Computer Engineering, Graduate School of Ajou University Suwon, Republic of Korea {kimjaebum, daeuky, youngko}@ajou.ac.kr Abstract—Content Centric Networking, which has simple data delivery procedures without using IP addresses, can improve content dissemination in wired networks and wireless Mobile Ad Hoc Networks (MANETs). However, the performance of CCN can be constrained in MANETs due to broadcast storming during content dissemination. Moreover, multiple content providers existing in MANET increase content announcement overhead. To solve these problems, we propose a TOPology aware CCN protocol (TOP-CCN). TOP-CCN employs the Multiple Point Relay (MPR) based packet flooding, the Publisher MPR (PMPR) based merged content announcement, and the expected hop count based flooding range control algorithm to reduce multiple content announcement overhead and broadcast storming problem. The simulation results show that TOP-CCN outperforms not only the pure CCN but also its variation in terms of content availability, network overhead, and content downloading time. (Abstract) Keywords— Content Centric Networking; Content Announcement; Broadcast Storming; MANETs (key words)

I.

INTRODUCTION

Multimedia data and consumer-generated data traffic are rapidly growing day by day due to the increasing population of smart, mobile devices. Despite these advancements in multimedia applications, the traditional IP-based Internet protocols are still utilized to service these overflowing amounts of data transactions. The problem is that the current Internet protocols do not effectively handle content requests from customers due to the address-content binding process overhead such as domain name service, end-to-end session based data delivery architecture [1]. In order to overcome this problem, several Information Centric Networking (ICN) architectures have been proposed in recent years. The Content Centric Networking (CCN) [2] is one of the famous ICN approach that does not use the IP address and endto-end sessions for data dissemination. When a consumer node wants to get a content item, the node broadcasts an interest packet containing the content name. If a provider node has the requested content item, it transmits the content item as a reply to the requester through the incoming face1 of the interest packet. Otherwise, it rebroadcasts this interest packet to search for another node that may possess the content item. Throughout these content request and delivery procedures, intermediate nodes between the content provider node(i.e., source node of the content item) and the consumer node are able to temporally store (i.e., cache) the delivered content item to reduce the path length and the downloading time. To ensure

these advantages in wireless networks, CCN over MANET has been demonstrated in a few works of recent years [3-6]. In tactical and emergency MANET, every node can generate unique contents such as surveillance video, voice, emergency message, etc. Thus, every node can be a content provider. To advertise their contents, each node generates considerable amounts of content announcement messages and content discovery messages, which in turn causes broadcast storming and duplication in packet transmission. However, various existing works in CCN cannot exploit these problems to support more efficient data dissemination in these MANET environments. In this paper, we propose a novel topology aware CCN, called TOP-CCN to support multiple content providers in MANETs. TOP-CCN employs Multiple Point Relay (MPR) based packet forwarding, publisher MPR (PMPR) based proactive content discovery, and flooding range control algorithms based on 1-hop and 2-hop neighbor information. In TOP-CCN algorithm, every node periodically elects its MPR, and PMPR through 1-hop broadcast content announcement message. Multi-hop content announcement message from the selected PMPR is broadcasted through MPR to reduce the number of flooding messages of the network. Based on a simulation study using ndnSIM [13], we show the improved performances of our proposal in terms of content availability, network overhead, and content downloading time, compared to other related works. II. BACKGROUND: CONTENT CENTRIC NETWORKING The CCN has two packet types of interest and content [2]. Also, it defines three types of data structure: the Forwarding Information Base (FIB), the Pending Interest Table (PIT), and the Content Store (CS). The FIB has a content prefix of content provider and the list of outgoing faces that are configured by a content prefix announcement from the content provider to forward interest packets. The PIT keeps track of the forwarded interest packet towards the content provider. By using the PIT, a content packet can be delivered to the content requester node called a consumer. Whenever an intermediate node receives a content packet, it stores the packet in its CS. In CCN, the interest and content forwarding process consists of content discovery, content delivery and content caching. A. Content discovery If a consumer wants to receive a content item, the consumer broadcasts interest packet with the content name that contains

1. The notation of the ‘face’ includes legacy network interfaces as well as the interface with application(s) in the node [2].

978-1-4799-2084-6/13/$31.00 ©2013 IEEE

ICON 2013

prefix of the content provider. In addition, the outgoing face information of the content provider is utilized to broadcast interest packet. If the node has not any content prefix and outgoing information of the content item in its FIB, then the interest packet is discarded [2]. The information of the FIB can be updated by various approaches such as proactive content announcement, reactive content discovery schemes, or manual setup by network managers [2, 7]. B. Content delivery and caching Content packet simply follows the chain of PIT entries to the requester node. Upon content packet arrival, each node lookups the content name of content packet using longestprefix matching [2]. If the node already has the same content packet in its CS, the packet is discarded. If the node has outgoing face information of the received content packet in its PIT, the content packet will be rebroadcasted through face information. In addition, the content packet is stored to CS of the node to ensure content cache effect. CCN provides innetwork caching functionality to achieve efficient content delivery [8]. In the CCN, every node has own CS to reduce the path length between the content provider and each consumer nodes. If a node receives a content packet from the neighbor, then content is stored in its CS using a caching policy that considers content popularity, freshness, etc [2]. III. RELATED WORKS The CCN can be a promising solution for content dissemination not only in wired networks but also in wireless networks such as MANET. In MANET environments, network robustness is dynamically changed due to interference, fading, and node mobility. Thus, end-to-end session based data dissemination and location or IP address based content path is frequently disconnected. The advantages of the CCN over MANET are as follow. First advantage is a mobility support without IP re-assigning. In traditional IP based wireless network, a mobile node that moves from a contended gateway another gateway must change or re-obtain new IP address to avoid IP conflict problem. Thus, IP address and end-to-end session based data communication services suffer from route disconnections. The CCN eliminates the complexity of IP reassignment and disconnection problems in mobile wireless networks by removing IP address. The other advantage is the reduced transmission overhead by in-network caching described in section II-C. To ensure these advantages in MANET, a few works have been presented in recent years. Soon Y. Oh designed a CCN for tactical and emergency MANETs [5], called MANET CCN. To solve broadcast storming problem, MANET CCN utilizes the overhearing mechanism. When a node overhears a neighbor’s interest packet, then the node aggregates the interest information. In addition, content reply mechanisms help to avoid packet collision. There are two types of messages, such as Reply and Request. A provider sends a Reply in response to the interest packet, and waits for a Request from the consumer before delivering the content. In MANET CCN, content items can be classified as situation based (e.g., temporal and spatial) items and topic based content items. Situation content items have higher priority than topic content items because they includes

location and time information. Thus, situation content items are quickly disseminated even though there are no request. M. Varvello et al. [7] analyzed the content availability performance and the routing cost of the various design of CCN over MANET approaches such as reactive, proactive and geographical through analytical model. According to this study, the optimal performance of reactive flooding approaches has the best content availability performance and the highest routing overhead. However, this result does not consider node mobility, interference, traffic size, and multiple content provider environments. M. Amadeo et al. [6] present a CCN based MANET protocol named E-CHANET that performs reactive based routing, forwarding, and reliable transport functions. In contrast with above algorithms, E-CHANET tries to solve not only the broadcast storming problem but also the channel unreliability and disruption through the extended content request and delivery procedures of the pure CCN. In ECHANET uses counter-based suppression mechanism before the packet transmission. When a node hears the same packet that has the same content prefix or content item, then packet is discarded for eliminating duplicated transmission. However, the consumer node does not effectively receives content packet in mobile environment due to topology dynamicity of the node caused by node mobility, and eliminated duplication packet. In addition, E-CHANET considers only a fixed single content provider. It is unsuitable assumption for MANETs. P. Sharma et al. [3] proposed a MANET-aware CCN algorithm on Android smartphones using Wi-Fi and Optimized Link State Routing (OLSR). This algorithm focuses on the message dissemination efficiency in highly mobile environments. To provide the higher probability of message delivery, MANET-aware CCN selects MPRs as content forwarding delegate. MPR can provide higher and optimized reachability to other nodes. MANET-aware CCN is based on CCN overlays because it brings the MANET optimizations for neighbor node reachability. Authors implemented two CCN overlays over OLSR using Haggle [10]. For the first overlay, they kept Haggle’s native algorithm for message forwarding. However, they removed link sensing and neighbor discovery function from Haggle CCN because OLSR already provides these functions. For the second overlay, the first overlay is extended by proposed MPR-aware forwarding algorithm. However, MANET-aware CCN uses the IP address to maintain neighbor information. IV. THE PROPOSED SCHEME Our TOP-CCN is designed to reduce the broadcast storming problem and to enhance the stability of content delivery in IEEE 802.11 based MANETs. To ensure robust content discovery, the TOP-CCN utilizes a proactive MANET routing approach [7]. Every node in TOP-CCN periodically broadcasts a content announcement (CA) packet, which contains the prefix information of the neighbor and sender. Thus, every node periodically updates its FIB table, 1-hop neighbor and 2-hop neighbor information to improve the accuracy of the face information. Moreover, TOP-CCN reduces the amount of packet flooding by employing three

novel algorithms such as MPR, PMPR, and flooding range control. A. TOP-CCN packet types TOP-CCN extends several attributes of the pure CCN packet format, which are interest and content packets. In addition, we introduce a new CA packet as shown in Table 1. This new packet has nine fields such as content prefix, ID, type, neighbor prefix list (NBR-PL), neighbor list (NBR-L), neighbor type list (NBR-TL), HOP count, cache hit ratio, and CA type. Content prefix is the name of the content provider that is assigned by the network manager or the prefix assignment algorithm [2]. ID is unique identifier of the node such as MAC address. Type attribute represents the type of the relay policy of the node. Its default value is non-MPR, which means that the CCN packet cannot be rebroadcasted. If some node is elected as a MPR by neighbor(s), then the node can rebroadcast the received CCN packet, and its type attribute is set to MPR notifying the flooding policy of the node. NBR-L is a ID list of the neighbor(s) gathered by the received CA message from neighbor(s). NBR-TL represents the type list of neighbor(s) such as MPR or not. More details of each type will be described later in Section IV-C. NBR-PL is a list of content prefixes of the entire neighbors. HOP count is the forwarded count of the packet that is increased by each forwarding node. Every node periodically measures a cache hit ratio during a period of time t. Cache-hit ratio is shown as the content reply ratio among the entire requested content items. When a node has fresh or popular content items in its CS, then the cache-hit ratio will be measured as a high value. When a node receives a CA packet from one of its neighbors, then 1-hop neighbor table, 2-hop neighbor table, and FIB are being sequentially updated. CA type attribute represents the type of the flooding policy of the CA message such as 1-hop limit and

multi-hop flooding. Both interest and content packets of the TOP-CCN have three extended fields such as the ID, expected hop, and the hop count. ID is the unique identifier of the sender such as consumer or provider of the packet. The expected hop represents the distance between the consumer and the content provider. The expected hop in the interest packet and the content packet are updated by the sender and relay nodes to limit the flooding range of the packet. B. Topology information tables update To ensure stable content delivery and discovery in wireless MANET environments, TOP-CCN defines the two tables for maintaining 1-hop and 2-hop neighbor information, respectively. The 1-hop neighbor table contains ID and lifetime of the node, whereas the 2-hop neighbor table includes 2-hop neighbor ID, connected 1-hop neighbor ID, and lifetime of the 2-hop neighbor. Both table information are periodically updated from the information of the received CA packets. The lifetime of the neighbor information is periodically initialized by CA message from the neighbor. When a lifetime of neighbor is expired, then the relative 1-hop and 2-hop neighbor information is removed. In addition, face information in the FIB and PIT is removed to prevent any wrong content delivery. C. MPR selection to reduce packet flooding overhead In CSMA/CA based wireless MANET environments, broadcast transmissions may incur potential problems such as the congestion and collision due to the shared medium characteristics [4]. However, CCN only considers the broadcasting approach to transmit packets. To mitigate the collision and congestion problem in CCN over MANET, TOPCCN selects a MPR set that can disseminate its received packets such as interest, content, and multi-hop CA packets. To select MPR, every node periodically calculates MPRcost using equation (1),

TABLE 1. TOP-CCN PACKET TYPES

Content Announcement Packet Content prefix (n) ID (48bit) Type (1bit) Neighbor prefix list (n) Neighbor ID list (n) Neighbor type list (n) HOP count (8bit) Cache-hit ratio (8bit) CA type (1bit)

Content data packet Content name (n) Signature (n) Signed info (n) ID (48bit) Expected HOP (8bit) HOP count (8bit) Data payload (n)

Interest Packet Content Name (n) Selector (n) Nonce (n) ID (48bit) Expected HOP (8bit) HOP count (8bit)

Content Prefix of the node Node identifier (MAC address, etc.) MPR or not Content prefix list of the neighbor(s) ID list of the neighbor(s) Type list of the neighbor(s) Number of forwarded hops of the packet Cache-hit ratio of the node Flooding type (0:single-hop / 1:multi-hop) Cons: Consumer node, Prov: Content provider node

Full prefix of the content item Digest algorithm, witness, etc… Publisher ID, Key locator, etc... Node identifier of the content requester HOP distance between Cons and Prov. Number of forwarded hops of the packet Actual Data, length is variable Cons: Consumer node, Prov: Content provider node

Full prefix of the content item Order preference, publisher filter, etc. Duplication prevention code Node identifier of the consumer HOP distance between Cons and Prov. Number of forwarded hops of the packet

χ͙n): The size is variable depending on the number of neighbors and prefix

MPRcosti = (1 − α) ×2hRatioi + α × ChachehitRatioi (1)

(0 ‫أ‬α ‫)أ‬

where i is the neighbor number, 2hRatioi represents the ratio of the 2-hop neighbor connectivity ratio of the neighbori. For example, a node has six 2-hop neighbors and its neighbori has three neighbors, then 2hRatio of interfacei is 0.5. If a neighbori has the same number of neighbors as the number of 2-hop neighbors of the node, 2hRatioi is 1. CachehitRatioi is given by CA packet from the neighbori. We use an exponential weighted moving average (EWMA) method to reflect not only network connectivity but also cache effect in MPRcost [6]. When a node needs to remove or add an entry in the 1-hop and 2-hop neighbor tables, then the MPR selection procedure will be invoked. The MPR selection algorithm is shown as algorithm 1. MPR nodes are elected among 1-hop neighbors of the node. First, a node selects its MPR node that has the highest MPRcost between neighbors. After selection of the MPR, all 2-hop neighbors that are connected through MPR are removed from the 2-hop neighbor list. When a node find remaining 2-hop neighbor, then additional MPR nodes are elected until 2-hop neighbor list is empty. If the node does not have any remaining 2-hop neighbors, then the selected MPR

Fig. 2. An example of PMPR selection and aggregated prefix information of the multi-hop CA in TOP-CCN

Fig. 1. An example of MPR selection for reducing packet flooding of the CCN

list is notified through NBR-TL in the CA message. Thus, every node in the network can decide the node type, which are MPR or not through the NBR-TL of the CA message from the neighbor(s). Fig. 1 shows an example of the MPR selection using CA packets. In Fig. 1(a), node 0 selects MPR nodes 1 and 4 to cover its 2-hop neighbors of nodes 2, 3, 7, and C. Following the same reason, every node selects its own MPR set as Fig. 1(b) to (c), respectively. After MPR selection, the entire MPR set of the topology is selected as Fig. 1 (d). All broadcast messages are relayed only by the MPR nodes. Thus, broadcast storming and duplicated transmission can be far relieved. D. PMPR selection to reduce CA flooding overhead In TOP-CCN, a CA packet has two types of prefix information such as content prefix of the sender and the neighbor content prefix list of the sender. When a node receives a CA packet, the content prefix list and the neighbor prefix list in the CA are inserted to FIB with received face information. In addition, 1-hop and 2-hop neighbor tables are updated using NBR-L, NBR-TL, and NBR-PL in the CA as describe in Section IV.B. Thus, a neighbor who receives a CA packet from its 1-hop or multi-hop neighbors can get the several prefix information of the neighbors of the sender. However, the proactive CA flooding increases network overhead such as packet collision, and congestion [11]. To Algorithm 1 MPR selection for limited flooding CA, interest and content packet of the TOP-CCN MPR{n} = Set of MPR CN{n}= Set of connected neighbor by MPR 2HOPNBR{n} = set of 2hop neighbor node MPR={͑Ϩ } CN={͑Ϩ } Repeat MPR {n}˥ 1hop_neighbor that has the highest MPRcost among 1hop_neighbors CN{n}˥ connected neighbors by MPR{n} Remove CN{n} from 2HOPNBR{n} until size of 2HOPNBR{n} == Ϩ

reduce the number of CA packet flooding, TOP-CCN selects PMPR nodes who are allowed to broadcast multi-hop CA messages. If the flooding type attribute of the received CA has single-hop value as 0, then the CA is not rebroadcasted. Multihop CA is generated only by the PMPR who has the largest number of neighbors among 1-hop MPRs. If the MPR node does not receive its neighbors prefix information during t, then the MPR is changed as PMPR to notify neighbor(s) prefix information to the network. Thus, content prefix information of the entire content provider in the network can be announced by only few multi-hop CA messages generated by PMPR. The PMPR selecting algorithm is started by 1-hop and 2hop table information change, and deals with three cases as shown in algorithm 2. The first is the absolute case. If a MPR node cannot find any MPR neighbor in its 1-hop neighbor table, then the MPR node itself becomes a PMPR for broadcasting multi-hop CA packet to the network. The second is the recovery case. If the MPR node cannot find content prefixes of the neighbor(s) from received CA packets during 6 seconds, then MPR also becomes a PMPR to notify neighbor(s) prefixes. The third is the competition case. When a MPR finds neighbor MPR list from its 1-hop neighbor table, it compares the number of neighbors between MPR and MPR list. If the MPR has the largest number of neighbors than all the neighbors, the MPR node will become the PMPR. Fig. 2 shows an example of PMPR selection. In this figure, PMPR set is 4, C, and 5 because they have the largest number of neighbors between each neighbor MPRs, respectively. Thus, Algorithm 2 PMPR selection for minimized CA flooding MPRNBR{n} = set of the entire MPR type neighbors MPR = MPR type neighbor NofNeighbors = The number of neighbors of the current node CA flooding type of the node = MPR: single_hop:0 if (current node type is MPR || MPRNBR{n} ==͑Ϩ ) CA flooding type of the node=PMPR:multi_hop:1 then return; if (does not receive prefix information of neighbor until ԩt) CA flooding type of the node=PMPR:multi_hop:1 then return; Repeat MPR ˥ first neighbor in the MPRNBR{n} if (The number of neighbors of the MPR < NofNeighbors) Remove MPR from MPRNBR{n} else CA flooding type of the node=PMPR:multi_hop:1 then return; until MPRNBR {n} == Ϩ

only three nodes in the network generate multi-hop CA messages. In addition, a multi-hop CA message is only relayed through MPR nodes in the network. Thus, network overhead of the proactive content announcement can be reduced in TOPCCN. E. Flooding range control using expected hop distance Generally, CCN packets are flooded to the entire network using FIB and PIT information for reliable content requests and deliveries. However, the unlimited packet flooding induces high network congestion and collision in MANET environments due to the inaccuracy of the FIB and PIT information caused by node mobility [12]. To solve this problem, we suggest an expected hop based flooding range control algorithm. Expected hop distance between the packet sender and the destination such as content provider and consumer is measured by multi-hop CA messages from the PMPR. When a node receives a multi-hop CA message from its neighbor(s), the receiver node updates the content prefix, outgoing face, and the expected HOP count information in the FIB table. Thus, the expected hop attribute of the interest and content packets can be filled using expected hop information in the FIB. If the node receives an interest or content packet from its neighbor, then it compares the expected hop count value between FIB and received packet. If the node has the same or lower expected hop count, the packet can be flooded into the network. When the expected hop count of the packet is lower than its own hop count, the packet will be discarded. V. PERFORMANCE EVALUATION In this section, we present performance evaluation studies such as reduced content dissemination delay, control message overhead, and increased accuracy of the face information in the PIT and FIB. To evaluate the performance of TOP-CCN comparing with pure CCN [2] and E-CHANET [7] in MANET environments, we utilize the ndnSIM with ns-3 [13]. Simulation parameters are presented in Table 2. The content request rate (interest rate) of the consumer is fixed to 1,000 per second. In addition, each consumer requests different content items to multiple content providers in the network. A. Flow analysis using network simulator visualizer. In the first simulation scenario, we verify our implementation and performance of the proposed scheme TABLE 2. SIMULATION PARAMETERS

Parameter PHY Content Providers/Consumers Topology Size Velocity MPRcost parameters CA interval Transmission Range Simulation Time Number of Simulations Chunk size / Content size Num.of Contents popularity dynamicity Cache replacement policy

Value IEEE 802.11g 18Mb/s 16/16 nodes (32 nodes) 500 x 500 m. Random topology 1 to 4 m/s (Random waypoint) ԩt: 4 sec,͑Յ: 0.5 2 seconds 25 m (indoor model) 600 Seconds 50 times per each scenario 10 kB / 1 MB 100 Zipf’s distribution LRU

(a) Pure CCN over MANET

(b) TOP-CCN over MANET Figure 3. The number of flooding message (interest/data) comparison in single consumer scenario using ns-3 visualization tool

through ns-3 network visualizer that can show the packet flow of the entire network with real-time simulation [13]. We use the 8x4 grid topology for the simplicity. Content requester is node 31 and content provider is node 5. Fig. 3 shows the flow snapshot of flooding packets in each scheme, which are pure CCN in MANETs, TOP-CCN without flooding range control and TOP-CCN with flooding control. A red circle represents a CCN node and a green arrow shows a packet flow. In Fig. 3(a), packet flows of pure CCN should fill out entire network to find the content provider and to reply the requested content. However, TOP-CCN as Fig 3-(b) dramatically reduces the number of flow by exploiting MPR and PMPR strategy. In this figure, the almost edge nodes in the network do not rebroadcast interest and content packet. In addition, flooding range is effectively limited through the expected hop count measured by CA messages from the entire node. B. Content availability, overhead, and delay comparison. Content availability represents the accuracy of the face information, which is deeply depending on the velocity of the node. When a node receives an interest packet from one of its neighbors, the incoming face information is added to reply the corresponding content packet to sender. However, in wireless MANET environments, incoming face information is not stable due to the link disconnections. Fig. 4 shows the average content availability depending on the node mobility. In this figure, TOP-CCN shows the highest content availability through CA based topology information update. If a node does not receive any CA packet from the connected neighbor during 6 seconds, the corresponding face information of a neighbor is removed to prevent packet loss and transmission overhead. The performance of TOP-CCN with flooding range control algorithm is slightly decreased through the limited interest and content packet flooding. Fig. 5 shows the network wide

smallest delay until 5 consumers scenario. However, 10 and 16 consumers environments, TOP-CCN with flooding range control shows the lowest delay due to the increased content availability and reduced network overhead. VI. CONCLUSION

Fig. 4. Average content availability vs. velocity of the node.

The future of CCN is very promising, as it is becoming the next-generation networking technology for efficient content dissemination in MANETs. Anticipating this trend, we present a topology aware CCN routing algorithm for MANETs. Our algorithm does not need TCP/IP stack to manage the topology and neighbor information. In addition, the suggested proactive based content discovery approach is used to improve content availability. Performance evaluation is conducted using ndnSIM with ns-3. The simulation result shows that our proposal incurs more decreased flooding overhead, decreased content delivery time and increased content availability compared to other schemes. In the future, we will optimize TOP-CCN with CA interval control and content priority based rate control algorithms. ACKNOWLEDGMENT

Fig. 5. Request message (interest and content announcement) flooding overhead of the network in 4 m/s velocity environment.

This research was supported by the MSIP, Korea, under the ITRC support program supervised by the NIPA (NIPA-2013(H0301-13-2003) and Basic Science Research Program through the NRF funded by the Ministry of Education (2012R1A1B3003573). REFERENCES [1] [2] [3] [4] [5] [6]

Fig. 6. Average content downloading time vs. consumers in 4 m/s velocity environment.

[7] [8]

overhead of the request message. In this figure, CCN over MANET shows the 8 Mb/s due to the uncontrolled message flooding. E-CHANET shows the highest performance until five consumers due to the overhearing and interest aggregation algorithm. However, its performance decreases in high congestion environment due to inaccuracy of the content availability. Thus, interest retransmission is dramatically increased in E-CHANET. Even if TOP-CCN uses the proactive CA approach, it shows the lowest overhead as 2.4 Mb/s in 15 consumer/provider environments due to the increased content availability. The average content downloading time is shown in Fig. 6. The performance is measured by a summation of 10 packet’s delay because a content item is fragmented into 10 packets. The delay performance of the E-CHANET has the

[9] [10]

[11]

[12] [13]

B. Ahlgren, et al., “A Survey of Information-Centric Networking,” Communications Magazine, Vol.50, pp.26-36, Jul. 2012. V. Jacobson, et al., “Networking Named Content,” CoNEXT, Dec. 2009. P. Sharma, et al., “A Case for MANET-aware Content Centric Networking of Smartphones,” WoWMoM, IEEE, Jun. 2012. M. Meisel, et al., “Ad hoc Networking via Named Data,” MobiArch, Sep. 2010. S.Y. Oh, et al., “Content Centric Networking in Tactical and Emergency MANETs,” WD, Oct. 2010. M. Amadeo, et al., “E-CHANET: Routing, Forwarding and Transport in Information-Centric Multihop Wireless Networks,” Computer Communications, Vol.36, pp.792-803, Apr. 2013. M. Varvello, et al., “On the Design of Content-Centric MANETs,” WONS, Jan. 2011. K. Katsaros, et al., “MultiCache: An Overlay Architecture for Information-Centric Networking." Computer Networks Vol.55, pp.936947, Mar. 2011. G. Tyson, et al., “Towards an Information-Centric Delay-Tolerant Network." NOMEN, Apr. 2013. E. Nordström, et al., “A Search-based Network Architecture for Content-Oriented and Opportunistic Communication,” Uppsala University Technical Report, 2009. M. Meisel, et al., “Listen first, broadcast later: Topology-agnostic forwarding under high dynamics,” Annual Conference of International Technology Alliance in Network and Information Science, Sep. 2010. C. Anastasiades, et al., “Content Discovery in Opportunistic ContentCentric Networks,” LCN, Oct. 2012. A. Afanasyev, et al., “ndnSIM: NDN Simulator for NS-3,” University of California Technical Report, 2012