9350
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 66, NO. 10, OCTOBER 2017
LOMCF: Forwarding and Caching in Named Data Networking Based MANETs Rana Asif Rehman, Member, IEEE, and Byung-Seo Kim
Abstract—Named data networking (NDN) resolves traditional transmission control protocol/internet protocol (TCP/IP)-based Internet problems (i.e., location dependent, complex usage, scalability, poor resource utilization, etc.) and is considered as an eligible candidate for futuristic Internet paradigm. In NDN-based mobile ad hoc networks (MANETs), the participating nodes are operated in highly dynamic and challengeable environment such as low battery power, channel fluctuations, intermittent connectivity, etc. Due to the broadcast nature of the wireless channel, the NDN-based MANETs highlight severe issues (e.g., packet collisions, flooding, data redundancy, and packet retransmissions), which further degrade network performance. In this paper, to cope with these problems, we have proposed a novel protocol, named location-aware on-demand multipath caching and forwarding for NDN-based MANETs. Performance of the proposed protocol is evaluated by using a simulator called ndnSIM. Extensive experiments along their results show that proposed protocol performs better as compared with the other recent proposed protocols in terms of content retrieval time, Interest retransmissions, and the total number of Interest packets injected, as well as discarded, in the network. Index Terms—Caching, content, information centric, interest, mobile ad hoc networks (MANETs), named data networking.
I. INTRODUCTION URRENTLY, Internet architecture is based on the TCP/IP paradigm in which communication is held using a device or machine physical address. It is hard to update the current Internet architecture and further enhance its communication parameters, which is also called ossification of the Internet [1]. Furthermore, the TCP/IP-based Internet has several drawbacks (e.g., location dependent, security, inefficient resource utilization). Currently, the quantity of wireless devices (e.g., smartphones, laptops, tabs) has increased sharply due to the rapid
C
Manuscript received June 27, 2016; revised December 19, 2016 and March 16, 2017; accepted April 27, 2017. Date of publication May 2, 2017; date of current version October 13, 2017. This work was supported in part by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2015R1D1A1A01059186) and in part by the Ministry of Education and NRF through the Human Resource Training Project for Regional Innovation (no. 2014H1C1A1066943). The review of this paper was coordinated by Dr. Y. Song. (Corresponding author: Byung-Seo Kim.) R. A. Rehman is with the Department of Computer Science, National University of Computer and Emerging Sciences, Chiniot-Faisalabad Campus, 35400 Pakistan (e-mail:
[email protected]). B. S. Kim is with the Department of Computer and Information Communication Engineering, Hongik University, Sejong City 30016, South Korea (e-mail:
[email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TVT.2017.2700335
, Member, IEEE
Fig. 1.
NDN hourglass.
development in technology. Consequently, Internet traffic has also increased exponentially, as stated by Cisco [2] and Google [3] in their reports. Today, users are more interested in getting/sharing their desired contents (e.g., video, music, photo, text) than the locations from where these contents are actually retrieved [4]. To fulfill the future application requirements, peerto-peer (P2P) and content delivery networks (CDNs) have been used as overlay networks [5]. However, these solutions are now considered inappropriate for the futuristic applications demand perspective and also have many drawbacks. Information-centric networks (ICNs) [6] is an emerging concept for future Internet architecture, in which content-based communication does not consider the device’s physical address. It provides the efficient content’s name-based communication functionalities for all the networks (e.g., mesh network, LAN, delay tolerant network, sensor and mobile ad-hoc networks) [7]. PURSUIT [8], 4WARD [9], COMET [10], CCN [11], CONVERGENCE [12], COAST [13], GreenICN [14], NDN [15], RIFE [16], CBMEN [17], UMOBILE [18], ENCODERS [19], BONVOYAGE [20], ICEMAN [21], POINT [22], and I-CAN [23] are some of the active projects being worked on in this arena. Named data networking (NDN) [15] paradigm is gaining more popularity and is widely accepted in the research community due to its simple communication architecture. Fig. 1 shows the hourglass architecture of both NDN- (right side) and TCP/IP (left side)-based paradigms. A mobile ad hoc network (MANET) is a combination of self-organized, resourceconstrained digital devices that can directly communicate with each other using a short transmission range. In this network,
0018-9545 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
REHMAN AND KIM: LOMCF: FORWARDING AND CACHING IN NAMED DATA NETWORKING BASED MANETS
9351
3) It introduces the distance-based caching policy that reduces the data packets duplication in the network. 4) The proposed protocol enhances the networks performance by considering the node’s remaining energy in its packet-forwarding mechanism. 5) The performance of the proposed protocol is evaluated using a well-known simulator, i.e., ndnSIM [30], which provides an official simulation environment for NDN-based networks. The rest of the paper is formatted as follows. Section II provides the description about the communication process in conventional NDN. In Section III, the previous related work is discussed. Section IV elaborates upon the description of the proposed LOMCF protocol in detail. In Section V, extensive simulation results are described, and conclusions are provided in Section VI. II. NDN IN A NUTSHELL Fig. 2.
Mobile ad hoc networks.
the topology and interconnectivity are unpredictable, as nodes randomly move, as well as dynamically appear and disappear, from the network [24], as shown in Fig. 2. Basically, a major type of communication in a MANET is data-centric in nature, such as situation awareness information, software updates, surveillance data, and so on [25]. For more than a decade, traditional IP-based protocols have been proposed for MANETs that are host-centric in nature. Moreover, by utilizing the TCP/IP approach, participating mobile nodes need to sustain end-to-end route connectivity to communicate. This phenomenon is also hard to attain in dynamic and lossy networks in which channels are intermittent and uncertain (e.g., MANETs). Similar to wired networks, NDN also shows fruitful advantages in wireless ad hoc networks [26], [27] and cognitive radio ad hoc networks [28], [29]. Due to the broadcast nature of the wireless channel, the NDN-based MANETs highlight severe issues (e.g., packet collisions, flooding, data redundancy, packet retransmissions) that further degrade the networks performance. Previously, the nodes utilized the in-networking caching property of the NDN that can store data packets at any node (i.e., leave copy everywhere caching policy). Consequently, the data redundancy is also increased in the network. In this paper, we propose a novel protocol, location-aware on-demand multipath caching and forwarding (LOMCF) for NDN-based MANETs. It is assumed that all the participating nodes can access their current location information at any time using any external service such as the global positioning system (GPS). The primary contributions of this study are highlighted as follows: 1) A reactive and on-demand forwarding protocol that takes a node’s current location into account, in its packetforwarding process, is proposed. 2) The proposed LOMCF protocol follows the multipath forwarding mechanism that mitigates the content retrieval time as well as increases the reliability of both Interest and Data packets.
NDN provides various functionalities, such as in-networking caching, security primitives, and multicast delivery. In NDN, communication is based on content names, and each content is uniquely identified by its name. Two types of packets, i.e., Interest and Data packets, are used in its communication procedure. Moreover, three types of tables, i.e., content store (CS), pending Interest table (PIT), and forwarding information base (FIB), are consumed for the manipulation of the Interest and Data packets. In CS, the Data packets are cached for future usage. If some other neighbor nodes require the same Data packet in future, they can fetch it from nearest node that has Data packet in its CS. On the other hand, PIT table is used for the manipulation of Interest packets. It contains all the pending request entries that other nodes are requested for particular Data packets. However, the FIB table is utilized for forwarding purposes. It contains the incoming/outgoing interfaces information and populated by using a routing protocol that periodically updates it. A simple communication process (i.e., step by step) in conventional NDN is illustrated in Fig. 3 in which a consumer node wants to download a Data packet from the provider node. First, it sends an Interest packet encapsulating the name of the desired content. When a relay node receives the Interest packet, it examines its CS table for the desired content. If a Data packet is available, the relay node sends it back to the consumer node. If it is not in the CS, the relay node checks its PIT table. If the same entry is available, it means some another node(s) already requested the same Data packet. As a result, the relay node discards the Interest packet. If there is no entry, it further examines the FIB table. If there is no corresponding FIB entry, then the relay discards the current transmission. Otherwise, it adds the current Interest packet entry in the PIT table and further forwards it to the provider node. Upon reception of the Interest packet, the provider node replies with the Data packet. The Data packet simply follows the PIT table entries in the reverse route. Every relay node can also store the Data packet in its CS table. In the future, if a neighbor node requires the same data packet, it will be provided by the relay nodes. This phenomenon reduces the content download time for future Interest packet requests.
9352
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 66, NO. 10, OCTOBER 2017
Fig. 3.
Conventional NDN communication process.
Unlike conventional NDN, some authors as in [31], first check the PIT table instead of CS table in order to reduce the lookup overhead. However, there is no clear consensus related to which approach is better one. Recently, many authors [32], [33] have still used the NDN conventional communication process, as described in Fig. 3. III. RELATED WORK Previously, different opportunistic [24] and content-based [34] models were used to achieve similar goals as NDN. However, NDN has gained considerable popularity in the wireless arena due to its simple communication paradigm. The authors in [35] present an NDN-based video streaming approach in wireless ad hoc networks. The authors claim that the NDN approach shows efficient quality of service (QoS) as well as productive results in a resource-constrained MANET environment. Raaid et al., in [36], compare the performance of both NDN-based and IP-based MANETs. Their study shows that NDN-based MANETs are considered a good candidate for emerging recovery application and natural disaster scenarios. In [37], the authors describe a new scheme (i.e., E-CHANET) for multihop wireless ad hoc networks. The E-CHANET scheme utilizes an extra data structure (e.g., provider table) during its communication process. Moreover, the E-CHANET scheme is proposed only for low-mobility scenarios and does not take into account the remaining energy of the nodes. Fabio et al. present a novel packet forwarding scheme that is based on bloom filters [38]. In this scheme, beacon information is periodically utilized and exchanged among participant nodes, which causes additional overhead in the network. In [39], the authors propose two schemes, packet collision avoidance and Interest packet aggregation for emerging and
tactical MANETs. In these schemes, two extra packets (e.g., request and reply) in addition to conventional NDN packets are used, which cause extra network overhead. In a previous study, the authors [40] utilized the NDN approach in wireless networks to reduce the video packet loss. They also introduce a new timeout algorithm for the measurement of variations in route-trip time estimation. Yu-Ting et al., in [41], elaborated upon a novel scheme (i.e., NAIF) to avoid the extra packet flooding in the network. In this scheme, a packet is forwarded or dropped based on the distance information and the data download rate. In a previous study, the authors [42] presented a scheme to download the content from mobile cloud. In this approach, the forwarder node is selected in each quadrant. Two extra packets (i.e., CMD, ACK) are also utilized in this process. Dabin et al., in [43], describe a broadcasting protocol for NDN-based wireless ad hoc networks. Additional packets (i.e., EFS, EFS-ACK) are periodically updated with the forwarding information on each node. Moreover, in this protocol, the hop distance value and link quality metric determine the eligible forwarder node. The authors in [44] propose a new unicast based protocol (i.e., AIRDrop) for the efficient content delivery in NDN-based wireless ad hoc networks. The AIRDrop protocol utilizes additional buffers and data structures, which drive extra burden in the network. In [45], two novel forwarding protocols, blind forwarding (BF) and provider-aware forwarding (PAF), are presented by the authors. The BF protocol describes the controlled flooding mechanism and reduces the Interest and Data packet flooding in the network. On the other hand, the PAF protocol utilizes a distance-based forwarding approach that is based on the hop count. In addition, both protocols do not consider the remaining energy of the nodes during their communication process. However, none of above mentioned schemes take the node’s remaining energy concept into account in their forwarding process. Most previously proposed schemes utilize the leave copy everywhere (LCE) caching policy. As a result, the number of duplicate Data packets is also increased in the network. IV. PROPOSED LOMCF PROTOCOL A. Problem Description Fig. 4 illustrates an example scenario of the NDN-based MANETs, in which mobile nodes can freely move anywhere. If a consumer node needs some content from a provider node, it follows the conventional NDN communication procedure and broadcasts an Interest packet to the provider node. Upon reception of the Interest packet, relay node 2 checks its CS for the availability of a Data packet. If it has data, relay node 2 sends it back to the consumer node. Otherwise, it further checks its PIT table. If there is no entry available, relay node 2 adds the Interest packet entry in the PIT table and further rebroadcasts the Interest packet toward other nodes. Similar to node 2, the other relay nodes (i.e., node 1 and node 4) adopt the same process of Interest packet forwarding. When node 2 rebroadcasts the Interest packet, the provider node as well as node 5 and node 6 also receive the same Interest packet. Nodes 5 and 6 (which do not have Data packets) further rebroadcast the Interest packet to other nodes using the same process as discussed before. In
REHMAN AND KIM: LOMCF: FORWARDING AND CACHING IN NAMED DATA NETWORKING BASED MANETS
Fig. 4.
9353
Problem scenario.
addition, nodes 8–10 also follow the same forwarding process. Hence, there is unnecessary Interest packet flooding in the network. The numbers in parentheses in the figure represent the sequence numbers of both Interest and Data packets issued in the network. Upon reception of the Interest packet, the provider node replies with the Data packet. When the Data packet comes back toward the consumer node, it simply follows the pending requests entries previously stored in the PIT tables of the relay nodes. When the Data packet is received by relay node 2, it checks its PIT table for a corresponding entry. If there is an entry, the relay node 2 forwards the Data packet toward the consumer node. It also removes the entry from the PIT table. Due to the Interest packet flooding, the similar request entries are stored on multiple nodes. Upon reception of the Data packet, the other relay nodes (e.g., nodes 5, 6, 8, and 9) also rebroadcast the requested Data packets. Consequently, there is unnecessary Data packet flooding in the network. However, the communication is only between a consumer and a provider node. Flooding packets is the hot research issue in NDN-based MANETs and lowers the network performance due to unnecessary packet transmissions. Moreover, in MANETs, the participating nodes have limited resources (e.g., battery power). Due to packet flooding in NDNbased MANETs, a participating node frequently takes part in the communication and rapidly consumes its remaining energy. As a result, the node dies soon, which further degrades the overall network performance. In the literature, the hop countbased provider-aware forwarding approaches were introduced in which the nodes forward the packets based on the hop distance from the provider node. However, these approaches do not perform well, especially in high-mobility cases. This is due to the fact that it is hard to sustain route connectivity, especially in uncertain and dynamic environments (e.g., MANETs). In NDNbased MANETs, each node has the capability to cache the Data packets. Previously, in most of the schemes, the nodes utilize the
same caching policy (LCE) that increases the data duplication in the network. For example, in above figure, nodes 6, 9, and 10 unnecessarily store the same Data packets although they are near the provider node. B. LOMCF Protocol Description We present a novel protocol, the LOMCF for NDN-based MANETs. The proposed LOMCF protocol considers the locations of both a consumer and a provider during its forwarding mechanism. A relay node forwards the Interest or Data packet only if it has less distance towards the consumer or provider node, respectively. This mechanism reduces the unnecessary packet flooding in the network. In addition, the forwarding process also considers the node’s remaining energy to improve the performance of the nodes. The LOMCF protocol utilizes the multipath forwarding process that has two benefits: First, high reliability is accomplished for transmitting both Interest and Data packets. Second, Data packets follow multiple routes. As a result, content is widely distributed over the network, which mitigates the content retrieval time for future requests. The LOMCF protocol also introduces a distance-based caching policy that takes into account the Data packet duplication in the network. According to this policy, it stores Data packets near the consumer compared to the provider node. Moreover, the LOMCF protocol also mitigates the Interest and Data packet collisions probability using a timer-based forwarding process. In the LOMCF protocol, a relay node lies into two modes based on its location and remaining energy. A relay node is in a potential Interest forwarder (PIF) mode if Dcp Eth
(1)
where Dpp and Dcp denote the distance from a previous relay node to a provider node and from a current relay node to a
9354
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 66, NO. 10, OCTOBER 2017
provider node, respectively. The distance is calculated using a well-known Euclidean distance formula as follows: Dij = (xj − xi )2 + (yj − yi )2 (2) where the xi and yi represent the position coordinates of the respective node i. In Fig. 4, node 2 is closer to the provider node (Dcp ) than the consumer node (Dpp ). Moreover, if it has higher remaining energy, then it is considered in PIF mode. Er shows the remaining energy of the node, and Eth represents the energy threshold. As explained in Section V-A, after extensive simulations, the value of Eth is set to 13% for evaluating the proposed protocol. If Er < Eth , then the node is considered in the Critical State. When a node is in the Critical State, it does not further send the Interest packets. Instead, it focuses on Data packet transmissions and satisfies its PIT table’s entries. A relay node is in a potential Data forwarder (PDF) mode and caches Data packets if Dcc
