Publisher Mobility Support in Content Centric Networks

Transcription

1 Publisher Mobility Support in Content Centric Networks Dookyoon Han, Munyoung Lee, Kideok Cho, Ted Taekyoung Kwon, and Yanghee Choi School of Computer Science and Engineering Seoul National University, Seoul, Korea {dkhan, mylee, and {tkkwon, Abstract As the high speed Internet and the smart devices become prevalent, it is expected that the traffic volume of the video conferencing/broadcasting on the mobile devices is increasing explosively. One of the promising solutions for the traffic explosion problem is the content centric networking (CCN), which solves the problem by focusing on the content instead of the location. Even though CCN can provide better delivery efficiency, robustness, security, and native subscriber mobility, little attention is paid to the publisher mobility problem, where the mobile devices produce the real-time traffic. In this paper, we propose PMC, a publisher mobility support protocol in CCN. By introducing two names and corresponding operations at the routers, PMC supports publisher mobility for the mobile video conferencing and broadcasting without violating the original CCN architecture. We believe that PMC is a validated and secured scheme following the philosophy of the CCN architecture to support mobile publisher. Index Terms Information centric networking, Content centric networking, Publisher Mobility, Mobile Broadcasting I. INTRODUCTION As the high speed Internet and the smart devices become prevalent, the traffic volume of the Internet is tremendously increasing, expecting to reach 1.3 zettabytes per year by 2016 [1]. To solve this traffic explosion problem and to deliver the contents more efficiently, there have been efforts to redesign the Internet architecture based on the content-oriented paradigm [2] [4]. Recently, content centric networking (CCN) architecture [2] is gaining momentum in the research community and the industry due to its merits. CCN can provide the enhanced content delivery by utilizing in-network storage and name-based routing. In the mobile environment, the increasing number of mobile devices also worsens the mobile traffic explosion problem. According to the CISCO report [5], mobile data traffic is expected to reach exabytes per month by 2017 and the largest portion of this mobile traffic is generated from the content delivery for the video streaming service [6]. Thus, CCN is expected as an effective solution to address this mobile traffic explosion as well as traffic explosion in the wired environment. To use CCN for the mobile environment, the mobility support is a crucial issue in order to provide stable mobile application service. The mobility support is classified as the publisher mobility and the subscriber mobility. In CCN, the subscriber mobility is easily supported by nature. The subscriber who has moved to another location and still wants to download the same content file simply reissues the interest packets. Then the interest packets are routed towards the publisher and the requested content will be forwarded to the subscriber. However, supporting the publisher mobility is not a trivial problem, since the routing table should be changed. Main challenges to support mobility of publihser can be summarized as follows: Fast FIB establishment: To forward interest packets properly to the publisher who moved to a new location, the forwarding information base (FIB) in the routers along the path should be updated properly and quickly. For the mobility support, we cannot rely on the ordinary routing protocols (e.g., OSPF, RIP) due to the timing constraints. If the application between publisher and subscriber requires less delay than the routing table convergence time or pulisher moves continuously before the routing table is converged, subscriber may not be able to download the contents. Maintaining a valid path to the mobile publisher: As the mobile publisher changes her/his Point of Attachment (PoA) continuously, a new path should be updated in the network so that a subscriber can download the publisher s contents. For this, temporal entry should be inserted in the FIB to forward the interest to the mobile publisher. Since the publisher moves frequently, the unused FIB entry should be expired to prevent the FIB table overflow. Also, we should maintain a path to the publisher regardless of the frequent PoA changes. In this paper, we propose PMC, a publisher mobility support protocol in CCN. By introducing two names and corresponding operations at the routers, PMC provides publisher mobility without violating the original CCN architecture. The main contributions of this paper are as follows: We investigate a mobility support, which does not violate the philosophy of the CCN architecture. In our proposed scheme, the location information is not used and there is no central entity dealing with mobility. Also, the updating process for the FIB information utilizes the interest and data packets which are basic components of CCN architecture. Our PMC supports a fast FIB update to maintain a valid path to the mobile publisher. Thus, the subscriber s qualty of experience (QoE) requirement can be satisfied /14/$ IEEE 214 ICOIN 2014

2 (a) Interest packet (b) Data packet Fig. 1. The CCN operation in comparison with the routing table convergence time of the original CCN architecture. The remainder of this paper is organized as follows: Section II describes the related work. In Section III, we explain the components and operations of PMC. Section IV evaluates the performance of our proposed scheme. Finally, this paper is concluded with future work in Section V. II. RELATED WORK In this section, we first review the concept and operation of CCN briefly, and then explain the existing mobility management techniques under CCN. A. Content Centric Networking (CCN) Content centric networking (CCN) [2], which is known as a representative architecture of information centric networking (ICN), can be one of the promising solutions for the future Internet. In CCN, communication is initiated by sending a request message (called an interest packet). Since CCN adopts a name-based routing, the interest packet is routed by a content name instead of a host address. The overall operation of CCN shown in Fig. 1 is described as follows [2], [7]: (1) when a requestor tries to get a content, she/he sends an interest packet with the name of the requested content; (2) when an intermediate router receives the interest packet, it looks up its content store (CS) to check the existence of the requested content. If the requested content is cached in the CS, the router sends the data packet to the requestor. Otherwise, the intermediate router forwards this interest packet to the next interface by looking up the content name in its forwarding information base (FIB); (3) after forwarding the interest packet, the forwarding history is recorded in the pending interest table (PIT), where each entry contains the content name and interface number from which the corresponding interest packet came. With the PIT, if a router receives the interest packets for the same content name, the router add their incoming interface number to the matching PIT entry without forwarding; (4) once the interest packet reaches the node that has the requested content, the node sends a data packet to the end host. This data packet is delivered following the backward path in the PIT; (5) during the data delivery, intermediate routers store the data packet in its CS in order to deal with future requests. Using this innetwork storage, CCN can provide efficient data delivery while reducing the redundant traffic for the same content request. B. Mobility Management in CCN As described in Section I, the mobility support is classified as the publisher mobility or the subscriber mobility. First, the subscriber mobility is easily supported by nature of the CCN architecture. When a subscriber has moved to other location and established a new connectivity, it reissues the interest packets to continue the download. In recent ICN researches, there are some efforts to enhance the subscriber mobility [8] [10]. Vasilakos et al. [8] proposes a proactive caching scheme for the fast content retrieval. In their scheme, some neighbor nodes are selected and contents are stored 215

3 (a) Mobility report (Interest packet) (b) Mobility report ACK (Data packet) Fig. 2. The forwarding decision of PMC components according to the report type in advance for the future mobility. Han et al. [9] propose a rate adaptation scheme to support stable video streaming service using multiple interfaces. Also, Ravishankar et al. [10] introduce a proxy agent for the effective mobility management. In their approach, the proxy agent deals with the content requests and content delivery instead of a mobile agent. In contrast with the subscriber mobility, supporting the publisher mobility is not a trivial problem since the routing table of all the relevant routers should be updated. To address this problem, Jiang et al. [11] add a locator information to the interest packet and rely on a mapping system that maintains the mapping of content name and locator information. In their scheme, a publisher updates her/his location information to the DNS-like mapping system. Then, an interest packet first visits the mapping system to know the publisher s location and it is forwarded towards the publisher. This approach can easily manage the publisher s new location. However, bypassing the entire interest packets to the centralized entity may generate a significant overhead. Moreover, main merits of CCN which came from the name-based routing disappear since the locator and identifier are tightly coupled. Kim et al. [12] also suggests an indirection approach. In their work, mobile nodes change their hierarchical content name whenever the mobility event occurs. Also, a DNS-like naming server, which is called rendezvous server, is proposed to manage these names. Then, subscribers who want to retrieve the content send an interest packet to the publisher after getting the content name of the publisher from the rendezvous server. However, this approach may cause some issues such as indirection delay and single point of failure. III. PMC DESIGN In this section, we explain the design of PMC, a publisher mobility support protocol in CCN. A. PMC Components We define two components for mobility support in CCN. PMC Router is a router that performs the PMC protocol. To support the publisher mobility, PMC routers maintain two types of entries in the forwarding information base (FIB), which are original entry and mobility entry. The original entry is an ordinary FIB entry for the stable and long-lived destination (e.g., a publisher in the wired environment) while the mobility entry maintains the next hop information for the mobile and temporal destination. The original entries are maintained by the routing protocol and the timer value is relatively long while the mobility entries are maintained by the PMC protocol and the timer value is relatively small. The original and mobility entries for the same prefix can be integrated as a single entry in the FIB with the priority. Since both entries within the FIB means that the publisher moved from the home to other places, the mobility entry should have higher priority than the original entry to forward the interest packets to the newly moved location. To handle the mobility, we define the home router (HR) of a publisher which announces the original entry of the publisher to the network. Therefore, when the publisher does not move, all interest packets are forwarded to the publisher through the home router. Mobile Publisher is a mobile device that produces the (real-time) content. To create and maintain the mobility 216

4 (a) When the publisher moves from the home to a PoA Fig. 3. The operation of PMC (b) When the publisher moves from a PoA to another PoA entry for the subscribers, a mobile publisher uses two reserved names: (publisher URL)/mobilityReport/Home and (publisher URL)/mobilityReport/PrevPoA. These names are used to report the publisher s mobility event to the home and the previous point of attachment (PoA) location. The publisher sends an interest with the name mobilityreport/home to keep the home updated with the path information to the publisher. On the other hand, the publisher sends an interest with the name mobilityreport/prevpoa to make a path from the previous PoA to the new PoA and to receive the missing interest packets arrived at the previous PoA while the publisher is moving. B. PMC Operation The overall operation of the PMC is depicted in Fig. 3. The operation is divided into two cases: (i) when the publisher moves from the home to a PoA and (ii) when the publisher moves from a PoA to another PoA. According to the report type (interest packet or data packet), each PMC router performs the forwarding and the FIB update as described in Fig. 2. From the home to a PoA: as shown in Fig. 3(a), (1) when the publisher moves from the home (i.e., router A) to a new PoA (i.e., router C), (2) it sends an interest with the name /snu.kr/mobilityreport/prevpoa to inform the home of its mobility (we assume that the publisher s URL is /snu.kr). Since the original entries for the /snu.kr are already established in router A, B, and C, the interest packet is forwarded to the home of the publisher easily. (3) On receiving the interest with the mobility report, the router A sends the Data to establish the mobility entries along the returning path. After forwarding back the Data towards the publisher, the intermediate routers (i.e., router A, B, and C) will update their FIBs by adding the next hop face with the highest priority in the entry for /snu.kr. (4) When a subscriber (or an end host) tries to download a content file from the mobile publisher, it will send an interest towards the publisher. Since router B updated the entry for /snu.kr with the mobility information, router B forwards the interest to router C, which in turn forward it to the publisher. By tracing the PIT entries, the Data from the publisher will be successfully forwarded to the subscriber. From a PoA to another PoA: Fig. 3(b) describes the operation when the publisher moves to another PoA, again. (5) In this example, the publisher moves from a PoA to a new PoA (i.e., router E). (6) Since the publisher moved from a PoA, it sends interest packets to the home and the previous PoA (We assume that the publisher can perceive whether the previous PoA is home or not). Note that the interest packets will be forwarded to router C by default since the FIB was updated with the mobility entries. To make the interest with mobilityreport/home to be forwarded to the home, we specify the face selection criteria in the Selector field in the interest packet (to follow the face with the second priority). Then, mobilityreport/home and mobilityreport/prevpoa interest packets are forwarded to the home and router C, respectively. (7) Router A and C send the Data to establish the new path. If there are buffered interest packets during the publisher s mobility, router C forwards the interest packets to the publisher using the updated path for the fast handoff. (8) The following interest packets will be forwarded using the updated path. IV. EVALUATION In this subsection, we evaluate the performance of PMC, focusing on the operation for the mobility support, described in Section III. The evaluation setting is as follows: Performance Metrics: we use (i) the chunk delivery delay and (ii) the data retrieval rate as metrics. 217

5 4500 Subscriber PoA2 PoA Management Server Home Router Mobile Publisher Fig. 4. Simulation topology Fig. 5. The comparison of chunk delivery delay Comparison Schemes: we compare PMC with (i) a Pure- CCN (relying on a periodic FIB update) and (ii) an indirection approach (using a DNS-like mobility management server). A. Simulation Environment We conducted simulations using a discrete event-driven simulator. As shown in Fig. 4, a network topology with 11 PMC routers is used for the simulation. A mobile publisher moves from the home router to the PoA2 via the PoA1 and a subscriber issues interest packets to obtain the contents from the mobile publisher. Also, there is a DNS-like matching server for an indirection approach. The simulation parameters are summarized in Table I. TABLE I SIMULATION PARAMETERS (ADOPTS FROM [13], [14]) Parameter CS size FIB size PIT timer Interest packet size Chunk(content) size Request interval Value 10 GB 10,000 entries 3,000 ms 125 byte 2,500 KB 3,000 ms In the above scenario, we compare the three schemes: (i) a Pure-CCN, (ii) an indirection approach, and (iii) PMC. In the Pure-CCN, the publisher s mobility event is reflected by periodic FIB updates. Since there is no explicit action to deal with the publisher mobility, the network slowly perceives the publisher s mobility event. Thus, interest packets cannot be forwarded to the publisher s new location until the next FIB update period. In the indirection approach, every packet bypasses the specific node, called management server (DNSlike mobility management server). A publisher reports her/his mobility event to the matching server to inform the subscribers about a new location of the publisher. In PMC, the publisher mobility is managed as described in Section III. B. Chunk Delivery Delay The chunk delivery delay is requestor s waiting time from issuing an interest packet to receiving a data packet. Fig. 5 shows the delivery delay of each content chunk. If a content request cannot be delivered to the publisher due to the publisher s movement, a subscriber reissues the interest packets. In this retransmission case, the delivery delay is calculated as the time difference between an issuing time of the initial interest packet and a retrieval time of the data packet. In Fig. 5, PMC shows shorter handover delay than Pure- CCN in terms of chunk delivery delay. To newly update the FIB table, PMC takes about 140 ms while Pure-CCN takes about 3,120 ms after the mobility event occurs. This implies that PMC can meet typical QoE service such as VoIP which needs a delay requirement of ms. Also, the indirection approach shows stable delivery delay under mobility conditions. However, the overall delivery delay of indirection approach is higher than that of Pure-CCN and PMC. Because every packet should bypass the DNS-like management server to get the publisher information, the indirection approach takes additional delay to detour the server. In summary, PMC achieves stable and shorter delay than other schemes since it does not detour the management server. C. Chunk Retrieval Rate Fig. 6 shows the sequence number of the content that a subscriber receives from a mobile publisher. Comparing with the Pure-CCN, PMC shows stable retrieval rate although the publisher s mobility events occur. Since PMC can perceive a mobility event quickly as it occurs, the mobility event does not significantly affect the retrieval rate of the content subscriber. In contrast, the publisher mobility considerably decreases the retrieval rate of Pure-CCN due to the absence of mobility management scheme. In Pure-CCN, the publisher s correct information cannot be utilized until the next FIB update period, which increases the content delivery delay. The retrieval rate of indirection approach shows a similar tendency with that of PMC. Compared with PMC, the indirection approach also achieves stable retrieval rate. However, it shows lower retrieval rate than PMC due to the indirection delay caused from a bypassing. V. CONCLUSION In this paper, we propose PMC, a publisher mobility support protocol in CCN. To set up the path from the old location 218

6 [9] B. Han, X. Wang, N. Choi, T. T. Kwon, and Y. Choi, Amvs-ndn: Adaptive mobile video streaming and sharing in wireless named data networking, in Proc. of IEEE INFOCOM NOMEN workshop, [10] R. Ravindran, S. Lo, X. Zhang, and G. Wang, Supporting seamless mobility in named data networking, in Proc. of IEEE ICC, [11] X. Jiang, J. Bi, Y. Wang, P. Lin, and Z. Li, A content provider mobility solution of named data networking. in Proc. of IEEE ICNP, [12] D.-h. Kim, J.-h. Kim, Y.-s. Kim, H.-s. Yoon, and I. Yeom, Mobility support in content centric networks, in Proc. of ACM SIGCOMM ICN workshop, [13] PARC, ccnx project, 2011, [14] N. Rozhnova and S. Fdida, An effective hop-by-hop interest shaping mechanism for ccn communications, in Proc. of IEEE INFOCOM NOMEN workshop, Fig. 6. The comparison of chunk retrieval rate to the new location of the mobile publisher, the publisher sends interest packets with the reserved names. We also define the operations at the routers to update the FIB entries for the mobility support without violating the CCN architecture. Simulation results show that PMC outperforms pure CCN and indirection approach in terms of the adaptability, efficiency, and stability. As our future work, we will conduct evaluations under various scenarios where the publisher moves across the different ISPs. Also, we will investigate a fast flooding scheme to support mobility event. ACKNOWLEDGMENT This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2013R1A2A2A ), and by the Cisco University Research Program Fund, an advised fund of Silicon Valley Community Foundation. The ICT at Seoul National University provided research facilities for this study. REFERENCES [1] Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, , ns341/ns525/ns537/ns705/ns827/white paper c html, May [2] V. Jacobson, D. K. Smetters, J. D. Thornton, M. F. Plass, N. H. Briggs, and R. L. Braynard, Networking named content, in Proc. of ACM CoNEXT, [3] T. Koponen, M. Chawla, B.-G. Chun, A. Ermolinskiy, K. H. Kim, S. Shenker, and I. Stoica, A data-oriented (and beyond) network architecture, in Proc. of ACM SIGCOMM, [4] A. Zahemszky, A. Császár, P. Nikander, and C. E. Rothenberg, Exploring the pub/sub routing & forwarding space, in Proc. of IEEE ICC Workshop Future-Net, [5] Cisco Visual Networking Index: Forecast and methodology, , ns537/ns705/ns827/white paper c ns827 Networking Solutions White Paper.html, May [6] A. Rao, A. Legout, Y.-s. Lim, D. Towsley, C. Barakat, and W. Dabbous, Network characteristics of video streaming traffic, in Proc. of ACM CoRR, [7] A. Udugama, C. Goerg, and A. Timm-Giel, Tutorial: Future internet with information centric networks, [8] X. Vasilakos, V. A. Siris, G. C. Polyzos, and M. Pomonis, Proactive selective neighbor caching for enhancing mobility support in informationcentric networks, in Proc. of ACM SIGCOMM ICN workshop,