Anchor-Less Producer Mobility Management in Named Data Networking for Real-Time Multimedia
RResearch Article
Anchor-Less Producer Mobility Management in Named DataNetworking for Real-Time Multimedia
Inayat Ali and Huhnkuk Lim Korea Institute of Science and Technology Information (KISTI), Daejeon, Republic of Korea University of Science and Technology (UST), Daejeon, Republic of Korea
Correspondence should be addressed to Huhnkuk Lim; [email protected]
Received 1 February 2019; Revised 5 April 2019; Accepted 17 April 2019; Published 13 May 2019
Academic Editor: Yuh-Shyan ChenCopyright © 2019 Inayat Ali and Huhnkuk Lim. This is an open access article distributed under the Creative CommonsAttribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work isproperly cited.Information-centric networking (ICN) is one of the promising solutions that cater to the challenges of IP-based networking. ICNshifts the IP-based access model to a data-centric model. Named Data Networking (NDN) is a flexible ICN architecture, which isbased on content distribution considering data as the core entity rather than IP-based hosts. User-generated mobile contents forreal-time multimedia communication such as Internet telephony are very common these days and are increasing both in qualityand quantity. In NDN, producer mobility is one of the challenging problems to support uninterrupted real-time multimediacommunication and needs to be resolved for the adoption of NDN as future Internet architecture. We assert that mobile node’sfuture location prediction can aid in designing efficient anchor-less mobility management techniques. In this article, we show howlocation prediction techniques can be used to provide an anchor-less mobility management solution in order to ensure seamlesshandover of the producer during real-time multimedia communication. The results indicate that with a low level of locationprediction accuracy, our proposed methodology still profoundly reduces the total handover latency and round trip time withoutcreating network overhead.
1. Introduction
Mobile devices have transformed into multimedia com-puters owing to the multitude of mobile applications anddiverse features, such as cameras, messaging, online con-tent sharing, and online mobile gaming. These smartfeatures of mobile devices and the human needs of creatingand sharing contents in real-time have enabled mobiledevices to play the roles of content providers and contentconsumers at the same time. The exponential growth indata production and its need for dissemination is becominga cumbersome issue in current network architecture. Theexisting Internet model is not built to combat this growingproduction and dissemination of data. Moreover, manyattempts were made to withstand this issue, which resultedin peer-to-peer (P2P) overlays on the IP network andcontent distribution network (CDN) [1]. These attemptsare still very feeble to ensure smooth network operation. According to a prediction by CISCO Visual NetworkingIndex (VNI) [2], global traffic will inflate to 3.3 ZB per year,or 278 Exabytes per month by 2021, thus raising moredisputes for the current state of the practice network ar-chitecture and its protocols.Many proposals have been witnessed in recent years for ascalable and reliable future Internet architecture [3–6].Among these proposals, information-centric networking(ICN) has gained much attention. ICN is a data-orientednetwork, where the information is retrieved from the Internetby naming the data, not the end hosts. Data objects are in-dependent of location unlike the IP address in the traditionalnetwork. This novel Internet architecture has addressed thechallenges faced by the current Internet architecture in-cluding scalability, addressing, name resolution, security,privacy, routing, and mobility. Among the ICN approaches,Named Data Networking (NDN) [7] is a very active, agile, andenterprising one. NDN operation is based on Interest/Data
HindawiMobile Information SystemsVolume 2019, Article ID 3531567, 12 pageshttps://doi.org/10.1155/2019/3531567 ackets, where a consumer (data requester) sends an Interestpacket containing the name of the required data. The NDNForwarding Information Base (FIB) helps forward the Interestpacket towards the data named in the Interest packet andretrieve the data in one or more Data packets. NDN also usesin-network caching for faster data access; hence the data canbe retrieved from the producer or the cache of the NDNrouter in the network. Many research challenges in NDNincluding naming, name resolution, routing, security forlarge-scale data, and mobility need to be resolved for NDN towin the race of future Internet architecture. The number ofmobile devices and the data they generated have drasticallyincreased over time [8, 9]. Mobility management is becomingmore salient because of the device transformation into a smartgadget that rapidly creates and shares content in real time.Therefore, device mobility management needs much atten-tion from the research community. As a result of NDN designprinciples, consumer mobility is natively supported by NDNas a change in the physical location of the consumer does notaffect the NDN data plane. There is no need for signallingfrom network to heal from consumer mobility but only In-terest retransmission by the consumer for the data that havenot yet been received works well. Handling producer mobilityis a challenging task and needs to update the network dataplane to successfully recover from outages due to producermobility.A few producer mobility management schemes for ICNhave been proposed for real-time services. However, thereis still room to design sophisticated algorithms to handleproducer mobility efficiently. The work in [10, 11, 12] hasreactive approaches towards mobility management wherethe packets are redirected after the handover completes andnotification is received from the producer. These ap-proaches suffer from path stretch problem like in Mobile IPfor mobility management in IP networking. Moreover,these reactive approaches encounter more content retrievaldelay than proactive approaches. The concept of redirectingpackets toward the new location was first introduced in thetraditional IP network (Mobile IP [13]) to support mobility.However, in mobile IP, all the packets have to be redirectedby the old access point towards the new location until thecommunication session ends, thus causing path stretchproblem for the entire session. The techniques used in[14, 15] use a proactive caching approach to minimize theInterest retransmission during a handover. The scheme in[14] assumes that the request pattern is known and thefuture requested content is proactively pushed to thenetwork caches ahead of handover to satisfy the Interestduring the handover. However, the mechanism in [15]pushes the content to network cache based on contentpopularity. However, these approaches cannot be appliedin real-time communication, where the contents are pro-duced in real-time after the Interest packets are received.Proactive caching techniques assume that the data arealready extant and pushed to the network cache beforereceiving an Interest for those data, which is not the case inreal-time communication. Real-time data are the data thatare generated and delivered immediately after they arerequested. The data generated in Internet telephony, messaging, and online gaming are examples of real-timedata, and the above proactive caching techniques formobility management will not support these applications,which account for a significant amount of Internet traffictoday.To solve the problems associated with the abovetechniques, i.e., path stretch and long delay due to re-active approaches and lack of support for real-timecommunication in proactive caching approaches, we,in this article, propose a proactive mechanism that is thefirst attempt to exploit location prediction techniques foranchor-less producer mobility management in order toensure seamless handover of the producer during real-time multimedia communication. The proposed meth-odology is based on the prediction of the future locationof a moving producer and proactively redirecting theInterests to the new access point (nAP) before thehandover completes. The proposed methodology is thefirst to exploit location prediction techniques in NDN formobility management. Our scheme reduces both thehandover latency and the round trip time without causingnetwork overhead. The mechanism also does not sufferfrom the path stretch problem. Moreover, it does notassume that the data are already extant; instead, the dataare generated in real-time and forwarded after the In-terest packets are intelligently delivered to the producerafter the handover.The rest of the paper is structured as follows: In Section2, we briefly offer an insight into NDN mobility and relatedwork. In Section 3, we describe the proposed locationprediction-based mobility management methodology indetails. We provide a detailed discussion of the results andcomparison with two legacy schemes in Section 4. Finally,we conclude the paper in Section 5 and present futurework.
2. Insight into NDN Mobility
In this section, we give an overview of operational aspects ofNDN, provide problems regarding mobility, and presentsome related work on mobility management in NDN.
NDN provides a new communicationparadigm that revolves around the content/data. The dataare retrieved by data names rather than the IP addresses ofthe machines that host the data. Once the data are forwardedto the network, they are cached at network routers (in-network caching) to satisfy future requests for the same datawith small delays. This in-network caching results in fasterservice access and resists congestion. Two data structures areused by NDN routers to aid packets forwarding,i.e., Forwarding Information Base (FIB) and Pending In-terest Table (PIT). FIB is populated by routing protocols anddifferent forwarding strategies, and it is used to forwardInterest packets towards the data. PIT keeps the records ofinterfaces on which the Interest packets are received and thedata names requested by those Interests. PIT ensures that theData packets follow the reverse path of the Interest packets.2 Mobile Information Systems third data structure called Content Store (CS) is used atrouters to cache data for satisfying future Interests. Uponreceiving an Interest in the router if the requested data areavailable in the router cache, they will satisfy the Interestwithout forwarding it to the next hop based on FIB. TheInterest packet is forwarded to the next hop based on FIBentry if the data are not available in router cache. Beforeforwarding the Interest packets, its incoming interface anddata name is stored in the PIT for data forwarding. When aData packet is received at a router, it will be forwarded to theinterface in the PIT entry against this data name. The datawill be dropped if there is no entry against this data name inthe PIT. The operation in the NDN router is explained inFigure 1.Mobility is an essential feature of future Internettechnology. NDN supports mobility and reduces thecomplexities of mobility support in the IP-based network.Mobility can be of two types, i.e., consumer mobility andproducer mobility. NDN naturally supports consumermobility by Interest retransmission and in-networkcaching. If a consumer after requesting data moves toanother network, the response will be delivered to therouter in the previous location, which will be cachedthere. Consumer at the new network will retransmit theInterest for the same data. In which case, the intermediaterouters will immediately respond to the Interest. In casethe data are not available at any of the network caches, theInterest will hit the data producer again, and the data willbe retrieved with a longer delay. Unlike consumer mo-bility, producer mobility is a complicated problem. Thistype of mobility highly affects communication. If aproducer moves to another network, the Interests will notbe delivered to it unless network converges. This type ofmobility cannot be tackled with Interest retransmission,as the new name prefix of the producer will not be knownto any NDN router unless network converges. Thecommunication session halts for at least the sum ofhandover time and the network reconvergence time asshown in Figure 2. The Interest packets are discardedat the producer’s old point of attachment (oAP) (11 . NDN relies on Interest retransmissionto heal from network outages owing to mobility. However, Interest retransmission is not a scalable approach whenthere are a large number of mobility events; also the end-to-end latency due to mobility in this approach is veryhigh. Therefore, some techniques have been proposed sofar to provide more robust and efficient mobility man-agement schemes. The schemes in the literature can bedivided into three categories: rendezvous-based schemesare more like DNS system in the current IP-based net-working, where a consumer sends a query to a particularnode called rendezvous server to find the location of theproducer before sending an Interest. The rendezvousserver keeps track of the location of all nodes in thenetwork. These techniques are easy to implement, but thedrawback is that they cause extra-overhead to maintain thelocation information of mobile nodes. The rendezvous-based schemes need the early binding of names that affectthe NDN data naming. Anchor-based approaches used ananchor node that redirects the Interests to the producer atthe new location. The anchor node should always be keptinformed of the producer’s movement. These schemespossess a single relay point (single point of failure) andsuffer from the path stretch problem. Performance eval-uation of anchor and anchor-less approaches are given indetail in [10]. The anchor-less approach has no pre-specified node to handle mobility. This approach is hard toimplement, but the schemes following such approaches arevery efficient in terms of performance. The schemes in[16, 17] are anchor-based mobility management schemes.The schemes proposed in [10, 11] are anchor-less mobilitymanagement schemes and suffer from the inheritedproblems from their corresponding approaches. Theproposed methodology in this work is based on the pre-diction of the future location of a mobile node. Theselocation prediction techniques can be exploited to provideanchor-less mobility management that highly reduces thehandover latency and round trip time even with lowprediction accuracy.
3. Location Prediction-BasedMobility Management
Producer mobility causes degradation in the quality ofservice during real-time streaming. The services aredisrupted when a producer moves and goes out of theradio range of its oAP and connects to another accesspoint (nAP). The Interest packets are dropped at the oAPas the new hierarchical name of the producer is notupdated in the network information. The communicationsession halts until the routing update period expires andthe whole network converges with this new producername information as shown in Figure 2. In NDN, theInterest packets are forwarded using FIB entries that arebeing populated by routing protocols, while the Datapackets follow the traces of Interest packet stored in thePIT to reach the consumer. Hence, the producer is notable to deliver Data packets of its ongoing streaming fromits new location right after the handover completes,unless it receives an Interest packet at the new locationusing its new hierarchical name prefix. This drasticallyMobile Information Systems 3aximizes the handover latency and hence round-triptime (RTT).
We use two types of Interest packets toimplement the mobility management scheme efficiently. TheInterest types are as follows:(i) INTEREST_PU: Interest path update message is sentby a mobile node (producer) to its AP to notify it about its mobility and probable handover. Thepacket format is shown in Figure 3(a).(ii) INTEREST_RED: Interest redirection is an Interesttype in which the Interests are forwarded to theproducer at new location after the handover. Thepacket format is shown in Figure 3(b).INTEREST_RED packet changes the content namefrom . . . /oAP/ale1 to . . . /nAP/ale1 (since the oAP havepredicted the nAP) and is forwarded based on the
Packet loss during producer mobility800750700650600550500 S e q u e n c e n u m b e r
11 11 . . . . . . . . . . . Time (sec) . . . . . . . . . . . . Handoff latency without mobility management scheme
Figure
2: Handover latency without using any mobility management.
Contentstore (CS) Pending InterestTable (PIT) FIBInterest ForwardData Add incoming faceReturn or drop the interestDownstream UpstreamPending InterestTable (PIT)Contentstore (CS) DataDiscard dataCacheForwardLookup missLookup hit
Figure
1: Interest and Data forwarding at the NDN router.
The mobility management can becontrolled at the network or mobile node. However, a moreeasy and efficient way is to make use of both to developmore efficient mobility management techniques. Theproposed methodology manages mobility using co-operation from both the network and the mobile node(producer). The location prediction scheme comes intoaction when the received signal strength (RSS) drops belowthe threshold (th � −
77 dBm). As the RSS drops below theth, the future location of the producer is calculated by theproducer itself using the speed and direction of movementobtained from its sensors. The producer sends its newlocation to the oAP in an INTEREST_PU message asshown in Figure 4 (Step 3). The INTEREST_PU messagecontains the X and Y coordinates of the producer’s futurelocation. The pseudocode for this is given in Figure 5. Thedetails of our prediction model are given in the followingsubsection. The se-lection of nAP is the very crucial and important part of thisscheme. There are many possible ways to select nAP;however, we use the relative distances of the producer’spredicted future location from access points. The relativedistance of the producer from APs is used in this workbecause, in our simulations, all the APs and mobile nodesuse the same transmit and receive power. Moreover, we usedthe free-space propagation model where there are no pathlosses, and it assumes one clear line-of-sight path from thetransmitter to the receiver. In such settings, the relativedistance is the fair method to be considered and increase theprobability of selecting the right new AP. Upon receiving theINTEREST_PU message from the producer, the oAP findsthe nAP using the Euclidean distance formula as explainedin the pseudocode in Figure 6. Here, we assumed that all the fixed routers and access points know each other geographiclocation. To find the distance between producer and nAP at afuture time ( T f ), oAP should know the new location of theproducer at T f . There are many methods to predict thefuture location of a mobile node [18, 19, 20] with differentprediction accuracy ranging from 45% to 90%. Our pre-diction model used in this paper is straightforward where topredict the location at a time ( T f ), the producer finds itsspeed and direction of movement and calculates its futurelocation coordinates at the time ( T f ). The producer sends itsnew predicted coordinates to oAP in an INTEREST_PUmessage. The modern mobile devices come with lots ofadvanced hardware and software sensors like accelerometer,gyroscope, digital compass, GPS, and speedometer. Thesesensors can easily calculate the speed and direction ofmovement of the mobile device (producer). The producerconstantly checks its RSS after it starts a movement. Theconnection becomes unreliable after the signal strengthdrops than a predefined threshold (here th � −
77 dBm) andthe node starts searching for reliable signals from other APsfor handover. Before the handover, the producer gets itsspeed and direction from the sensors at that particular timeand calculates its future coordinates and sends them in anINTEREST_PU message to the oAP. The new speed at aparticular time is determined by V new � min max v old + Δ v, , v max , if p ≤ p s , , otherwise , ( ) where v old is the old speed obtained from the producerdevice sensors and Δ v is the change in speed after time ( t ), Δ v � v new − v old , and it is uniformly distributed on [ −
3, 3]km/hr. v max is the prespecified maximum speed (30 km/hr inour simulation) of a mobile node. p is the uniformly dis-tributed probability between [0, 1], and p s is thresholdprobability ( p s � ϕ new � ϕ old + Δ ϕ , ( ) where ϕ new is the new direction after time ( t ) and ϕ old is theold direction. Δ ϕ is the change in the direction and isuniformly distributed on [ − π /4 , π /4 ] . Euclidean distancebetween producer and AP is an efficient method to select Name X-coordinates(new predicted position) Y-coordinates(new predicted position) Nonce Interest_PU(value = 0) (a)
Name Selector(order preference, publisher filter etc..) Nonce Guiders(scope, interest life time) Interest_RED(value = 1) (b)
Figure
3: Interest packet types. (a) INTEREST_PU. (b) INTEREST_RED.
Mobile Information Systems 5 inding and sending future location coordinates to oAPPossible event: Received signal strength (RSS) drops than threshold (th = –77 dBm).(1) Case event (2) If RSS ≤ th:(3) Get speed ( v new ) and direction ( ϕ new )(4) Calculate future coordinates ( x p , y p ) using current speed and direction(5) Construct INTEREST_PU message(6) Send INTEREST_PU to oAP Figure
5: Pseudocode for path update message generation.
New AP (nAP) Selection by old AP (oAP) Possible events: = Arrival of INTEREST_PU message at oAP, ∂ = arrival of nAP new reachable prefix notification from producer at oAP (wrong prediction).(1) Case events(2) When :(3) Get predicted coordinates ( x p , y p ) from INTEREST_PU message(4) Predict nAP:(5) Dist i = (AP i , x – x p ) + (AP i,y – y p ) (6) AP i = min (Dist i )(7) nAP = AP i (8) Return nAP(9) Redirect Interests toward nAP(10) Store copies of Interests for small time ( t s )(12) if ∂ :(13) Redirect Interests again based on nAP prefix information in ∂ (14) Discard the stored Interests packets(15) else:(16) Do nothing(17) Discard the stored packets after time t s Figure
6: Pseudocode for new access point selection.
Rtr-1 Rtr-2Cons-1 ale 1ale 1oAPnAPAP1: Interest sent path before handover 2: Content received path before handover 3: RSS ≤ thINTEREST_PU sent(location update message sent to oAP)4: Find nAPRedirect Interests to nAP5: Redirected Interests received At producer’s new location6: Content receivedafter handoverNDN router 1 NDN router 2
Figure
4: Location prediction-based mobility management scheme. t s to redirectit later, in case, the nAP prediction goes wrong. Here t s isset a little more than the RTT in our simulation to ensurethe time required for nAP to notify the oAP after thehandover. After the handover, if the nAP does not receivean Interest packet, it will notify its new reachable prefix tothe oAP via a broadcast. If the oAP does not receive anynotification from the nAP, it will assume that the nAP waspredicted right and the Interest packets have successfullybeen redirected to it, or it will assume that the producerhas completely disconnected. In both cases, the storedInterest packets will be removed after time t s . The pseu-docode for nAP selection and redirecting the Intereststowards it is given in Figure 6. The nAP is anticipated tohave producer connected to it after successful handover atlayer 2. We have set the layer 2 handover delay ( L ) to100 ms. Layer 2 handover delay is the time it takes by anode to disassociate itself from one AP (e.g., oAP inFigure 7) and establish a connection with another AP(e.g., nAP1 in Figure 7). The redirected Interestsare forwarded in a separate Interest packet type calledINTEREST_RED. INTEREST_RED packet changes thecontent hierarchical name from . . . /oAP/ale1 to . . . /nAP/ale1 (since the oAP has predicted the nAP) and is for-warded based on the FIB entry corresponding to this newname. The intermediate NDN routers upon receiving theredirected Interest update the previous PIT entry for thatname prefix and also notify the FIB about the changes forupcoming Interests. The intermediate routers then forwardthe INTEREST_RED packets according to the longestprefix matching based on FIB towards the nAP.The INTEREST_RED packets after reaching the nAP are buffered for some time and then forwarded to theproducer immediately after the handover completes, andits connection establishes with the nAP. If the nAP pre-diction goes wrong, and the producer at the new locationdoes not receive the Interest packets, it publishes itsnew name prefix by broadcasting it. The oAP after receivingthis new name prefix assumes that the nAP is predictedwrong and forwards the Interests to this correct new lo-cation of the producer according to the longestprefix matching based on FIB. In this method, the pre-handover Interest forwarding to the nAP highly reduces thehandover latency and overall RTT from consumer to theproducer. The Interest packets path to the producer beforehandover is Consumer ⟶ AP ⟶ Rtr1 ⟶ Rtr2 ⟶ oAP ⟶ Producer (Figure 4). The Data packets followthe reverse path of the Interest packets. The path for theInterest packets that were in transit before updatingthe FIBs of the intermediate routers after handoveris Consumer ⟶ AP ⟶ Rtr1 ⟶ Rtr2 ⟶ oAP ⟶ Rtr2 ⟶ nAP ⟶ Producer. The Data packets path afterhandover will always be Producer ⟶ nAP ⟶ Rtr2 ⟶ Rtr1 ⟶ AP ⟶ Consumer. The scheme does not sufferfrom path stretch problem. Only the Interest packets thatwere sent before forwarding the redirected Interests sufferfrom path stretch.
4. Results and Discussion
We have simulated the following fat-tree backhaul net-work that resembles a real ISP topology as shown inFigure 8 using the NS-2 simulator. The scenario consists of4 mobile producers and two stationary consumers, whereeach consumer is continuously sending requests for data tothe two producers for the entire duration of the simulation.We have repeated the simulation 100 times, and the resultsare averaged over many runs. The producers randomlymove, and handover can occur to any adjacent cell. We
ProdoAP nAP2nAP1oAP Range nAP1 Range nAP2 RangeConnectedStrong RSSWeak RSS
Figure
7: Access point selection.
Mobile Information Systems 7ave 16 cells with full wireless coverage (WiFi 802.11n),where the producers and consumers are located in dif-ferent cells initially. The 2D view of cells and the initialposition of the nodes are shown in the figure. In theproposed location prediction-based mobility managementmethodology, the node’s future expected location is pre-dicted before finding the nAP. The oAP does not drop theincoming Interests during the handover time. Instead,oAP immediately forwards the Interests to nAP. HereRTT/2 ( ≈
25 ms) (delay experienced by INTEREST_REDfrom oAP to nAP) is completely eliminated by overlappingit with Layer 2 handover delay ( L ) of 100 ms as describedin Figure 9.It is clear that the Interest packets sent during the ( L ) handover delay time are delivered to the producer imme-diately after 100 ms. The Interest packets are not waiting forany notification from producer or network after thehandover. The following equation gives the total handoverlatency in this scheme: H p � L + α ,H p ≈ L , ( ) where α is very small and covers the propagation time ofInterests from nAP and producer after the handover and a little delay in predicting the actual handover incident. H p isthe total handover latency, and L is the layer 2 handoverdelay. The equation H p � L is the optimal mobility man-agement at the network layer as it eliminates the handoverdelay at the network layer. The handover latency we obtainedin our simulation is 109 ms for location prediction-basedmethodology during accurate prediction as compared to134 ms in zone flooding and 150 ms in Interest forwarding[12] as shown in Figure 10(a) and 10(b), respectively. Thebetter performance is because the Interest packets are for-warded to the new location (nAP) before the handovercompletes as the nAP is already predicted. In the Interestforwarding scheme, after the producer attached to the nAP,it sends an update message to oAP, and the oAP then CoreBackhaulEdgeAccess 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Prod-3 Cons-1Prod-1 Prod-4Prod-2 11 23 4 5 67 89 10 1612 13 14151 Cons-1Prod-2 Prod-4Prod-1 Prod-32D viewICN routerICN AP ConsumerProducer Cons-2Cons-2
Figure
8: Fat-tree backhaul network topology.
Layer 2 handover delay = 100ms1 25 100RTT/2Interests redirection Time in milliseconds
Figure
9: Overlapping of L handover delay and Interest re-direction delay. h ) in the locationprediction based scheme is given by the equation below andshown by the peak value in Figure 11. RTT is defined as thetotal time it takes to send an Interest packet and receive thecorresponding Data packet:RTT h � L + RTT t , ( ) S e q u e n c e n u m b e r L + α Handover latency = L + RTT/2 . . . . . . . . . . . Time (sec) . . . . . . . Handover latency with location prediction-based schemeHandover latency with zone forwarding scheme (a) S e q u e n c e n u m b e r L + α Handover latency = L + RTT
11 11 . . . . . . . . . . . Time (sec) . . . . . . . Handover latency with location prediction-based schemeHandover latency with interest forwarding scheme (b)
Figure
10: Handover latency in (a) location prediction and zone flooding scheme and (b) location prediction and interest forwardingscheme.
Mobile Information Systems 9here RTT t is the average round trip time between con-sumer and producer before handover. The RTT duringhandover is highly reduced to 139 ms, and RTT in Interestforwarding scheme is 160 ms, while in zone flooding, RTT is144 ms, almost similar to our scheme. However, the zoneflooding scheme creates a very high network overhead as it isbased on sending multiple copies of Interests to all APs in azone. Therefore the zone flooding scheme is not scalable. Onthe contrary, the location prediction-based scheme createsno network overhead. RTT due to path stretch in the Interestforwarding scheme is shown in Figure 11, while the locationprediction-based scheme does not suffer from the pathstretch problem.Figures 12(a) and 12(b) show average handover latencyin different schemes with varying percentage of accuracy inprediction. Since the prediction cannot be 100% right,Figure 12(a) shows that with a very low prediction accuracyof 50%, location prediction based method substantiallyreduces the average handover latency, yet keeping excellentcontent to the Interest ratio as shown in Figure 12(b). RTTand handover latency for zone flooding are quite small, butits content to the Interest ratio is also very small because ofits multicasting nature, and it cannot be regarded as thebest solution. Moreover, our proposed approach signifi-cantly reduces average handover latency, RTT, andmaintains an outstanding content to Interest ratio as ev-ident from the simulation results. On the contrary, Interest forwarding scheme has a good Content to Interest ratio, butit has big RTT, handover latency, and suffers from pathstretch.
5. Conclusion and Future Work
In NDN, the producer mobility is one of the strenuous andchallenging tasks during real-time streaming. The handoverlatency constitutes the delay caused by dissociation andreassociation of a mobile node to oAP and nAP, respectively( L delay), and the delay caused by network convergencetime. In this work, we have exploited location prediction forproducer mobility management to show its impact on thedesign of such mobility management techniques. The sim-ulation results have shown that using location predictiontechniques for mobility management can significantly re-duce the total handover latency to approximately equal tolayer 2 handover delay by eliminating the delay caused bynetwork convergence. With a minimal prediction accuracyof 50%, the methodology tends to perform better, whichshows the benefits of using location prediction for producermobility management. It also reduced end-to-end RTTwithout creating any network overhead. The results of thismobility management approach may turn out as the baseknowledge for the design of a sophisticated location pre-diction technique for mobility management. As of futurework, we plan to design an efficient location prediction Peak RTT in interestforwarding scheme Proposed location prediction-based schemePeak RTT for packets affected due to mobilityPeak RTT in zone flooding schemeRTT due to path stretch in interest forwarding schemeThere is no path stretch overhead in proposed scheme and zone flooding200405060708090100110120130140150160170 230 260 290 320 350 380Sequence number R o u n d t r i p t i m e ( m s )
410 440 470 500 530 560 590End-to-end RTT of location prediction-based schemeEnd-to-end RTT of interest forwarding schemeEnd-to-end RTT of zone flooding scheme
Figure
11: Round trip time comparison of the three mobility management schemes.
Table
1: Handover latency comparison.Scheme Handover latency RemarksInterest forwarding H p ≈ L + RTT RTT between producer old location and new location is addedwhich varies with topology and other network parametersZone flooding H p ≈ L + RTT/2 RTT/2 is added here which also varies as RTT variesLocation prediction based (proposed) H p ≈ L RTT is completely eliminated with our proactive approach (Figure 9)
10 Mobile Information Systemsechnique and apply it in our proposed methodology formobility management in NDN.
Data Availability
No data were used to support this work. Moreover, thesimulation code is available from the authors upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
References [1] L. Zhang, A. Afanasyev, J. Burke et al., “Named data net-working,”
ACM SIGCOMM Computer Communication Re-view , vol. 44, no. 3, pp. 66–73, 2014.[2] CISCO Visual Networking Index (VNI),
Cisco Visual Net-working Index: Forecast and Methodology, 2016–2021 , White Paper, CISCO Visual Networking Index (VNI), Cisco, TX,USA, 2017.[3] V. Jacobson, D. K. Smetters, J. D. Thornton, M. F. Plass,N. H. Briggs, and R. L. Braynard, “Networking named con-tent,” in
Proceedings of the 5th International Conference onEmerging Networking Experiments and Technologies , pp. 1–12,ACM, New York, USA, December 2009.[4] T. Koponen, M. Chawla, B.-G. Chun et al., “A data-oriented(and beyond) network architecture,”
ACM SIGCOMMComputer Communication Review
Publish/Subscribe forInternet: Valencia FIA book
European Commision , vol. 3, no. 3, pp. 34–36, 2010.[9] CISCO Visual Networking Index (VNI),
Cisco Visual Net-working Index: Forecast and Methodology, 2014–2019 , WhitePaper, CISCO Visual Networking Index (VNI), Cisco, TX,USA, 2015.[10] D.-H. Kim, J.-H. Kim, Y.-S. Kim, H.-S. Yoon, and I. Yeom,“End-to-end mobility support in content centric networks,”
International Journal of Communication Systems , vol. 28,no. 6, pp. 1151–1167, 2015.[11] J. Lee, S. Cho, and D. Kim, “Device mobility management incontent-centric networking,”
IEEE Communications Maga-zine , vol. 50, no. 12, pp. 28–34, 2012.[12] D. Han, M. Lee, K. Cho, T. Kwon, and Y. Choi, “Publishermobility support in content-centric networks,” in
Proceedingsof the International Conference on Information Networking2014 (ICOIN2014) , pp. 214–219, IEEE, Phuket, Thailand,February 2014.[13] C. E. Perkins, “Mobile IP,”
IEEE Communications Magazine ,vol. 35, no. 5, pp. 84–99, 1997.[14] H. Farahat and H. S. Hassanein, “Proactive caching forproducer mobility management in named data networks,” in
Proceedings of the IEEE 13th International Wireless Com-munications and Mobile Computing Conference (IWCMC) ,pp. 171–176, Valencia, Spain, June 2017.[15] T. Woo, H. Park, S. Jung, and T. Kwon, “Proactive neighborpushing for enhancing provider mobility support inContent-Centric Networking,” in
Proceedings of the 2014Sixth International Conference on Ubiquitous and FutureNetworks (ICUFN) , pp. 158–163, IEEE, Shanghai, China, July2014.[16] F. Hermans, E. Ngai, and P. Gunningberg, “Global sourcemobility in the content-centric networking architecture,” in
Proceedings of the 1st ACM workshop on Emerging Name-Oriented Mobile Networking Design—Architecture, Algo-rithms, and Applications (NoM ’12) , pp. 13–18, Hilton Head,SC, USA, June 2012.[17] Y. Luo, J. Eymann, and A. Timm-Giel, “Mobility support forcontent centric networking,”
Telecommunication Systems ,vol. 59, no. 2, pp. 271–288, 2015.[18] H. Abu-Ghazaleh and A. S. Alfa, “Application of mobilityprediction in wireless networks using markov renewal the-ory,”
IEEE Transactions on Vehicular Technology , vol. 59,no. 2, pp. 788–802, 2010.[19] N. Samaan and A. Karmouch, “A user centric mobilityprediction approach based on spatial conceptual maps,” in A v e r a g e h a n d o v e r l a t e n c y ( m s )
10 20 30 40 50 60 70 80 90 100nAP prediction accuracy (%)Location prediction-based schemeInterest forwarding schemeZone flooding scheme (a)
10 20 30 40 501.21.110.90.80.70.60.5 C t o l r a t i o Handover eventsLocation prediction-based scheme (50% predictionaccuracy)Location prediction-based scheme (0% predictionaccuracy)Interest forwarding schemeZone flooding scheme (b)
Figure
12: (a) Average handover latency. (b) Content to interestratio.
Mobile Information Systems 11 roceedings of the IEEE International Conference on Com-munication (ICC) , vol. 2, pp. 1413–1417, Seoul, Korea, May2005.[20] H.-W. Ferng, W.-Y. Kao, S. David, C.-L. Liu, H.-Y. Chen, andH.-Y. Gu, “A dynamic resource reservation scheme withmobility prediction for wireless multimedia networks,” in
Proceedings of the IEEE 60th Vehicular Technology Conference(VTC2004-Fall. 2004) , vol. 5, pp. 3506–3510, Atlantic City, NJ,USA, October 2004.
12 Mobile Information Systems omputer Games Technology
International Journal of
Journal of
Engineering
Volume 2018
Advances in
FuzzySystems
International Journal of
ReconfigurableComputing
Applied Computational Intelligence and Soft Computing
Advances in
Artificial Intelligence
Civil Engineering
Advances in
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Advances in
Multimedia
International Journal of
Biomedical Imaging
Engineering Mathematics
International Journal of
Robotics
Journal of
Computational Intelligence and Neuroscience
Mathematical Problems in Engineering
Modelling &Simulationin Engineering
The Scientific World Journal
Volume 2018
Human-ComputerInteraction
Advances in