ReView by River Valley This

Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

IET Research Journals

Submission Template for IET Research Journal Papers

Handover Performance Analysis for ISSN 1751-8644

doi: 0000000000

managing D2D mobility in 5G Cellular

www.ietdl.org

Networks
Kaouthar Ouali1∗ , Meriem Kassar2 , Kaouthar Sethom3
1
University Tunis El Manar, ENIT, SMS Research Unit, Tunisia
2
University Tunis El Manar, ENIT, Communication Systems Laboratory, Tunisia
3
University of Carthage, ENICarthage, SMS Research Unit, Tunisia
* E-mail: oualikaouthar@gmail.com

Abstract: Device-to-Device (D2D) communication is seen as an important emerging technology and a hot research area for 5G
cellular networks that takes advantage of the physical proximity of communicating devices. However, its implementation faces
some technical issues such as interference management, resource allocation, mobility management, security, etc. Recently, many
research works have tried to deal with those problems. In this paper, we focus on D2D mobility management issue and we give
a comparative study on the existing techniques in cellular networks managing the mobile devices connections and maintaining
service continuity when moving between coverage areas. Firstly, we describe the different D2D handover management techniques.
Secondly, through analyzing and categorizing a wide range of the latest research works on Distributed Mobility Management
(DMM), we select some approaches that could be suitable for D2D mobility management in 5G cellular networks. Finally, we
mathematically analyze the handover performance of the selected solutions, we compare them and we conclude that Software
Defined Networking (SDN) based solution outperforms the others in terms of handover delay, packet loss, ping-pong effect as well
as the handover failure probability.

1 Introduction Base stations are placed at near intersections where it will provide
the largest coverage to outdoor users. However, coverage can be
Traffic in mobile networks is continuously growing and the need for poor within indoor environment, as signals may be blocked by con-
more sophisticated services are pushing current researches to pro- crete and steel. Thus, poor indoor coverage can be solved by small
vide a tight integration between different wireless technologies, a cells deployment that is costly and sometimes not possible. Here,
high throughput, a low latency and a high capacity to manage mobil- D2D communication comes as a novel approach to overcome poor
ity. Thus, it requires a new generation of mobile communications, coverage conditions.
referred to as the 5G (5th Generation). 5G is the next genera- Basically, D2D communication implies that wireless proximate
tion of mobile technologies beyond LTE-A (Long Term Evolution- devices communicate in a peer-to-peer model without involving the
Advanced) technology [1]. It is needed to satisfy the future telecom- base stations or the core networks [7]. This mode is not totally seam-
munications trends such as the efficient use of the spectrum, the less to the mobile cellular networks and occurs either in licensed
diversity of services and applications, the enhanced user experience, spectrum (i.e. inband) or unlicensed spectrum (i.e. outband) [8, 9].
the usage of high frequency bands, the full duplex communication Currently, D2D communication technologies include Wi-Fi Direct
systems, the energy consumption saving, etc. It intends to handle the [10, 11], that connects two devices to a wireless network created
explosion of the number of connected devices efficiently. This new by one of them (≈ 50 meters), Bluetooth and NFC (Near Field
generation includes new and existing technologies like GSM (Global Communication). By using D2D communication, mobile devices in
System for Mobile communications) [2], UMTS (Universal Mobile proximity can freely form a network between them or within com-
Telecommunications System) [3], LTE (Long Term Evolution) [3], panies, regardless of the Internet. Also, mobile devices can freely
WiMax [4] and Wi-Fi [5], etc. move but keeping a certain transmission distance between them that
Like all the previous technologies and networks, 5G has to deal depends on the D2D technology being used. In fact, with licensed
with many challenges [6]. The capacity of handling a high number of spectrum, the D2D transmission distance is about 1 Km [12], much
connected devices is the most challenging issue for the proliferation longer than other technologies. Thus, it is likely to be used in many
of 5G technologies, according to communication industry analysts. ways.
Interoperability issue between SDN (Software Defined Networking) Although the usefulness of the D2D technology, there are plenty
and NFV (Network Function Virtualization) is also at the fore- of open challenges that still in their infancy and need to be solved in
front of discussions. Security, minimizing the latency, improving the D2D communication to become a successful paradigm in 5G cellular
throughput and saving energy consumption are still important issues. networks [13]. In fact, the use of D2D technologies risks to increase
Hence, many reflections must focus on the design of an architecture the energy consumption of mobile devices. In other words, any form
or a framework to solve these problems, each one with its solution of D2D communication, based on Wi-Fi and Bluetooth, requires two
but all integrated into a coherent sensible way. wireless interfaces and would have to run in parallel with cellular
In this context, one solution to the main problem of increasing radio connectivity, and thus affecting the devices batteries. Interfer-
capacity is the integration of several key technologies and the use of ence management is also a big challenge since D2D users, operating
HetNet (Heterogeneous Networks). HetNet architecture comprises within inband mode, can disrupt the cellular services. Security fea-
an overlay of a macrocell network with additional tiers of densely tures of D2D communication should be better studied as they are less
deployed small cells such as microcell, picocell, femtocell, relays robust than those used in cellular networks especially when using
and Wi-Fi access points. However, in a typical dense urban envi- Wi-Fi or another technology. Resource allocation, modulation as
ronment, cellular coverage can be spotty, specially for indoor users.

IET Research Journals, pp. 1–13

c The Institution of Engineering and Technology 2015 1

2018/07/30 17:04:37 IET Review Copy Only 2

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

well as channel measurement are to be further investigated. Mobil- the BS (Base Station) takes the decision to switch the D2D commu-
ity management is an interesting issue because the current handover nication infrastructure and initiates the handover process.
management procedures in the centralized mobile networks archi- The second type is the MD handover between base stations while
tecture can not provide a reliable D2D communication. Thereby, we engaging a D2D call with another device. Here, two procedures are
are investigating the Distributed Mobility Management (DMM). needed: mobility of a single device, called single mobility), mobility
In this context, many analysis, surveys, studies and tutorials were of the two devices together, called dual mobility (that can be half
proposed in the literature to overcome some D2D challenges: D2D or joint handover), where the links between MDs and BSs (control
interference and resource allocation [14], D2D taxonomy [9], D2D plane) can be switched to one or two new BSs.
mode selection [15], etc. In contrast, our paper gives a vision of how The third scenario is the vertical handover allowing service continu-
the D2D mobility can be treated efficiently in 5G cellular networks ity and ubiquitous coverage for mobile devices. The D2D commu-
through a comparison between existing DMM techniques. Moreover, nication must be maintained when MD (i.e single vertical handover)
it gives a handover performance analysis to adapt the D2D mobil- or two MDs (i.e. dual vertical handover) move from one network
ity management in 5G cellular networks. Such a work has not been attachment point to another of different technology (exp. the Wi-Fi
proposed in the literature. Direct to the 5G cellular network).
The paper is organized as follows: Section 2 describes the han-
dover management in the context of D2D communications. A brief
review of different DMM techniques is presented in Section 3. In
Section 4, we compare these DMM solutions with the SDN-based
DMM approach. Section 5 presents the analytical models for some
selected DMM solutions. Through analytical modeling based perfor-
mance evaluation, we show in Section 6 that SDN-based technology
is the most suitable one for an efficient D2D mobility management
in terms of handover latency, packet loss, etc. Finally, Section 7
concludes our work.

2 D2D Mobility Management

Mobility management is the process that maintains information
delivery at the location of MDs (Mobile Device) and manages
their connections when moving between coverage areas. It includes
two processes: location and handover management. The handover
management is classified in two different schemes: the client (host)-
based and network-based handover management [16]. In the first
scheme, MD is involved in the mobility procedure. It is required to
maintain its mobility session and informs its location changes which
leads to a modification in the MD IP stack. The second scheme is
totally different. MD is no longer involved in the signaling process
and is unaware of the mobility or handover procedure. Only network
components participate in the handover management process.
In the context of D2D communication, when MDs are going through
a D2D communication and when they are in proximity to each other,
they undergo a joint handover. The devices may also not be in prox-
imity or one of them may get handed over to another neighboring Fig. 1: D2D Mobility Scenarios
cell, resulting in a half handover. Hence, the base stations providing
the first and the second cells coordinate to determine resources for
D2D communication between the involved MDs.
When D2D users perform simple LTE HO (HandOver), several
problems might emerge such as high latency, extra resource wast- 2.2 D2D Handover Management Techniques
ing, extra signaling exchange, and interrupted D2D data link. Since
traditional and current mobility solutions as well as handover pro- 5G does not specify D2D handover procedures, and the current
cedures can not support the D2D service continuity requirements LTE handover procedures can not support the service continuity
and provide a reliable D2D communication. Hence, D2D mobility requirements. Hence, several drawbacks might emerge as specified
management is the most challenging issue that we investigate in this previously. Therefore, an efficient D2D handover mechanism is nec-
paper. essary to provide more reliable D2D communication and to maintain
the D2D service continuity. In order to overcome the challenges
induced by devices mobility in D2D communication, we are ded-
2.1 D2D Mobility Scenarios icated to study earlier works on mobility characteristics of D2D
mobile users and how they affect network performance. Here, we
The introduction of direct D2D communication in 5G cellular net- give a research work on handover management technique in a D2D
works leads to introduce a new mobility management procedure communication in order to select the most suitable mechanism that
(defining a new D2D handover management solution) in addition can best optimize D2D handover performance.
to the traditional cellular handover. However, mobility management
in D2D communication is more complicated since the transmission Yilmaz et al. [17] present a smart mobility management solution
range of a D2D communication is limited even if both the D2D trans- for D2D communications in 5G cellular networks aiming at reduc-
mitter and the receiver have the capability of mobility. Here, three ing control signaling and improving End-to-End (E2E) latency in
main types of handover are deduced for a D2D link as illustrated in a network-assisted D2D communication. In fact, the involvement
Figure 1. The first type, D2I (Device-to-Infrastructure) mobility, is of several BSs in the D2D radio control may degrade the QoS
the switching of the D2D communication between the conventional requirement in terms of latency due to large backhaul delay and may
cellular communication and the direct communication link. In fact, increase the signaling overhead due to the exchange of necessary
when the D2D link becomes inadequate for a communication (i.e. information between the BSs. Thereby, it is beneficial to keep the
bad QoE (Quality of Experience) or bad QoS (Quality of Service)), D2D pair controlled by the same BS, but using the regular handover

IET Research Journals, pp. 1–13

2 c The Institution of Engineering and Technology 2015

2018/07/30 17:04:37 IET Review Copy Only 3

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

conditions for each MD independently does not guarantee this. several SDN controllers (network controller, RAN controller, BS
The authors proposed the first solution ’D2D-aware handover’. This controller and device controller) as depicted in Figure 8(b). Each
latter postpones the handover of the D2D control from BS1 to BS2 controller is in charge of controlling and managing each part of the
until the signal quality of BS1 becomes less than a predefined thresh- cellular network. In fact, a handover procedure of a D2D communi-
old, in order to keep the D2D pair controlled by the same BS, and cation may involve both the MD and BS controllers as well as the
then decreases the signaling overhead. The second solution ’D2D- RAN controller. In the case of dual mobility where the two MDs
triggered handover’ is proposed to reduce the control signaling by move to the same BS, the RAN controller will not be included since
clustering many D2D pairs within a minimum number of cells or there is not a resource coordination between the BSs. In the D2D to
BSs. It selects the target cell for handing over the D2D control of the D2I mobility, the handover control is handled by both the RAN and
D2D cluster that controls the majority of the D2D members. Simula- the network controller as well as the BS controller to establish a new
tion results show that the proposed solutions can reduce the network route between the two MDs.
signaling overhead and improve the E2E latency of a D2D commu-
nication by maximizing the time period when the MDs are under the Orsino et al. [20] presented an eSDN-ProSe architecture where
control of the same small cell. This work does not take into account NFV and C-RAN functionalities are used to improve the users QoS
the signal quality of the D2D pairs nor discuss their mobility in dif- and QoE for a D2D communication. To offload the heavy traffic of
ferent directions and scenarios. macrocells in low power BSs and D2D communications, the authors
A D2D handover mechanism was proposed by Chen et al. [18] use the Cloud-Radio Access Network (C-RAN) with the joint use
including a handover decision method which consisting of several of SDN and NFV. Hence, a new D2D-aware C-RAN architecture
parameters such as Time to Trigger of D2D (T T TD ), the time has been developed to efficiently manage the establishment of the
interval that is required to satisfy a certain D2D signal quality. D2D links. To proof the applicability of the proposed architecture,
These variables are used for making handover decisions, such as the authors test the handover of the entire D2D users crossing an
Joint Handover procedure, Half Handover procedure, or even no overlapping zone. In fact, all the messages exchanged in the tradi-
handover. In Joint Handover, as shown in Figure 2(a), the two MDs tional handover mechanism (i.e. X2 protocol) between the source
(defined as User Equipment, UEs) are in close proximity so they can and target BS could be avoided and requested to the new Evolved
hand off simultaneously to the neighboring cell and provide service Packet Core (EPC). Furthermore, to alleviate the link connecting the
continuity. When the UEs are not in close proximity or only one EPC and the radio access of a BS, the authors propose to reroute the
UE have to hand over to another cell, a Half Handover procedure, messages related to the request of paths and bearer setting through
as shown in Figure 2(b), is proposed in which one UE hands over the core network. Through the flow diagram of the handover proce-
to the target cell and the other remains connected to the source cell dure, Orsino et al. had shown that their architecture is able to reduce
by maintaining service continuity. In a short period of time, the sec- the number of signaling messages during a D2D handover from 22
ond UE hands over seamlessly to the same cell. Figure 3 represents messages (LTE X2 handover procedure) to 9 messages as well as the
a hybrid handover procedure that combines both the joint and half handover latency.
handover, triggered by a D2D handover decision method to ensure Despite the huge work presented in this paper, the authors con-
that the target cell is the most appropriate to a D2D communication. sidered only the handover of the D2D pair from a source to a target
According to the simulation results, the authors compare between BS. The half handover is not considered. In order to more claim the
five procedures: LTE A3 HO scheme∗ , Yilmaz’s D2D-aware han- applicability of the architecture, validation and simulation should be
dover [17] scheme, the Joint Handover procedure (Joint scheme), performed.
the Half Handover procedure (Half scheme) and the hybrid scheme.
The proposed mechanism reduces the number of D2D HO and 3GPP also investigates D2D communications as Proximity Ser-
hence minimizes the D2D HO failure rate compared to the other vices (ProSe). In particular, different D2D mobility scenarios are
schemes. The D2D mode ratio also presents the greatest values with described in [21] and the mobility requirement in commercial net-
the proposed procedure because it enables UE in proximity to com- works are described in [22].
municate in a D2D mode for a long time by maintaining a good
QoS. The amount of LTE HO, D2D HO and mode switch (i.e. from Table 1 summarizes different works related to D2D mobility. If a
D2D/cellular mode to cellular/D2D mode) are lower in the proposed solution satisfies a given property we denote it by + (yes) or - (no).
scheme than the other procedures due to the fact that it presents The qualitative features ’Feasibility’ indicates that some solutions
the lowest D2D HO failure rate and the highest D2D mode ratio. can not be easily simulated and realized due to complexity of the
This leads to reduce the additional signalling caused by the mode architecture or due to increasing the exchanged messages. As we
switch, D2D communication re-establishment and interruption, etc. can see, only the techniques [17], and [20] are focusing on D2D
Although the good results that insure this handover mechanism, the handover performance metrics. The other works show only how their
authors consider only the quality of signals to decide a handover, proposed architectures can efficiently support D2D mobility.
they do not introduce all parameters for handover decision. In addi- Among those last described works, only the D2D handover mech-
tion, UEs do not follow a random movement but they move straight anism presented by Chen et al. [18] can be considered in our work
either at the same direction or away from each other. This approach since it presents a detailed and appropriate mechanism (Fig. 2) that
is also based on a centralized mobility management. we can rely on to make analysis in the following sections.

Yazici et al.[19] introduced a new programmable all-SDN 5G

network architecture featuring hierarchical control capability which 3 Distributed Mobility Management (DMM) for
supports D2D communication. More specifically, the authors pro- D2D Communication
posed a 5G C-Plane to provide a CMaaS (Connectivity Management
as a Service) approach in order to ensure the control of mobility and Current systems are based on a centralized and hierarchical struc-
the routing for devices in the network and to keep track on the mobil- ture, even in D2D communication, where multiple access networks
ity during the D2D communication phase. The proposed approach is are connected to a central mobility anchor, and packets to deliver are
unified due to the merge of both the Radio Access Network (RAN) always forwarded via this anchor. This may poses some scalability
and Core Network (CN) functions, which are implemented as net- and routing issues which can be overcame by removing the single
work applications running on one or more hierarchical controllers. anchor and adopting a distributed mobility management (DMM) in
In particular, it solves mobility management of MDs by deploying a flat network architecture. In this section, we focus only on the
distributed mobility management. A centralized versus distributed
comparison is well investigated in the paper proposed by Giust et al.
∗ The
[23].
LTE Event A3 is triggered when a neighbouring cell becomes better Distributed mobility management (DMM) is a recent approach
than the serving cell by an offset. discussed in IETF [24]. In this new paradigm, the concept of a flatter

IET Research Journals, pp. 1–13

c The Institution of Engineering and Technology 2015 3

2018/07/30 17:04:37 IET Review Copy Only 4

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

(a) D2D Joint Handover procedure (b) D2D Half Handover procedure

Fig. 2: D2D Handover procedure [18]

Table 1 D2D Mobility Management Techniques

Techniques D2D mobility D2D mobility type Architecture Feasibility HO performance
Yilmaz et al.[17] - Dual mobility 5G + E2E latency, Signaling overhead
Chen et al.[18] + Half and dual mobility LTE-A + D2D HO failure rate, D2D mode ratio
Yazici et al.[19] + Single, dual mobility and D2D to D2I 5G-SDN + -
Orsino et al.[20] + Dual mobility 5G-SDN - HO latency

3.1 IETF based DMM

Distributed Mobility Management paradigm used to design a flat

mobility architecture by introducing several anchors closer to the
users. The main families of DMM solutions are based on network-
based mobility and described in the paper propsoed by Giust et al
[27].

3.1.1 PMIPv6-based DMM Solution: This solution is based

on the Proxy Mobile IPv6 (PMIPv6) mobility protocol [27]. The
main idea of the PMIPv6 solution is to use one of the elements in net-
work as a proxy for the Mobile Node (MN) and the Care of Address
(CoA) will not be assigned to the MN rather the CoA will be with
the proxy of the MN. PMIPv6 architecture reveals the existence of
Fig. 3: D2D Hybrid Handover Procedure new entities comparing to the MIPv6 one. Indeed, the Correspon-
dent Node (CN) is responsible of sending the data to the MN which
is connected to the Access Router (AR). A Local Mobility Anchor
(LMA) which is the equivalent of the Home Agent (HA) in MIPv6
solution and a Mobile Access Gateway (MAG) keep track of the
MN. The domain where PMIPv6 is implemented is called PMIP
system is developed. Mobility management functions can be dis- domain or Localized Mobility Domain (LMD). The advantages of
tributed in the data plane, whereas the control plane may be either implementing a PMIPv6 mobility solution reside both on the MN
centralized or distributed, the anchor can be moved closer to the which can change the AR without changing its IP address and with-
device. Several DMM family solutions and technologies have been out any software or hardware modification (i.e. standard protocol
proposed in the IETF and 3GPP standards. However, it is not yet stack) and on the common network that is able to accommodate var-
clear whether these DMM solutions can be applied in a D2D com- ious access technologies (WiMAX, 3GPP, WLAN, etc.).
munication. In the following sections, we demonstrate the best can- With DMM, the centralized entity is removed and pushed in the
didate technology that can efficiently support D2D communication access node. So, the IP sessions have the gateway in the access net-
in 5G cellular networks as considered in [25, 26]. work. The data packets no longer need to pass through the packet

IET Research Journals, pp. 1–13

4 c The Institution of Engineering and Technology 2015

2018/07/30 17:04:37 IET Review Copy Only 5

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

core network. Hence, we present the mobility access router (DMM-

GW equivalent to the MAG) which is in charge of forwarding the
user traffic from and to the Internet and to handle the mobility of
MN while moving from one MAG to another. The traffic redirection
is coordinated with a node that we call Control Mobility Database
(CMD) (i.e the equivalent of the LMA reduced to only a control
plane entity) which is a store for users mobility sessions. This node
is not crossed by the user traffic as shown in Figure 4. The traffic is
routed through DMM-GW and the CMD is not involved.

Fig. 5: Routing-based DMM

attachment to the access network, by selecting the appropriate S-

GW (Serving GW) and P-GW (Packet GW) that are geographically
close to the UE’s access point as shown in Figure 6(a).
LIPA enables a UE connected to a small BS to access other local
IP area network without the operator’s core network be crossed by
the user plane as shown in Figure 6(b). In this scheme, a Local-
Fig. 4: PMIPv6-based DMM
Gateway (L-GW) is added which has a direct link with the femtocell
that acts as a GGSN (UMTS) or P-GW (LTE). When the LIPA
default bearer (LTE) or the Packet Data Protocol (PDP) context
The DMM-GW sends signaling messages, Proxy Binding Acknowl- (UMTS) is established, the data flows are directly rerouted to the L-
edgement (PBA), to the control entity and it gets back the informa- GW and from there into the local network without passing through
tion with IP parameters in Proxy Binding Update (PBU) that should the radio access network or the core network. Hence, LIPA is
be sent to MN in order to be able to connect to the network. At compatible with any device without any software modification.
this step, MN can communicate with any node in the network and Up to Release 10 [29] and specially in 3GPP specifications, both
the data packets go through the gateway and from the gateway to SIPTO and LIPA have a very limited mobility support. In other
the Internet. In case of a handover, MN will be attached to a differ- words, the bearer is lost when the mobile looses contact to the fem-
ent DMM-GW and the same signaling messages will be exchanged. tocell. But in Release 11, a work item on LIPA/SIPTO Mobility in
Indeed, the DMM-GW sends messages to the CMD which basically Local Network (LIMONET) [30] is presented to provide to those
distributes the information to the former and the new DMM-GW in mechanisms a limited mobility support. Indeed, SIPTO mobility is
order to make them capable to set up a tunnel to allow the commu- still limited to the case where the S-GW/P-GW are in the RAN level
nication. The new DMM-GW allocates then a new prefix to the MN. and it is not taken into account within the local network. LIPA mobil-
This latter will be able of keeping the former flow alive with its IP ity support is considered for handovers between femtocells managed
address and it uses the prefix recently affected for the new traffic. by the same L-GW.
3.1.2 Routing-based DMM Solution: Compared to the previ- 3.2.2 Adoption of PMIPv6 in the 3GPP EPC architecture:
ous one, this solution has different principles as depicted in Figure PMIPv6 has been considered to support IP mobility in LTE archi-
5 and in [27]. It uses the Border Gateway Protocol (BGP) to per- tecture [31]. In this study, to support PMIPv6 in LTE networks, the
form the mobility functions in order to deliver the packets to and P-GW is used as the LMA of PMIPv6 and the S-GW is used as the
from moving devices. It removes any anchors and lets the routing MAG. The X2 and S1 handovers are the same as conventional LTE.
mechanism reestablish the path. The DMM-GWs are involved with PMIPv6-LTE uses Generic Routing Encapsulation (GRE) tunnel-
all the BGP routers in the network domain. For initial attachment, ing between S-GW and P-GW instead of GPRS Tunneling Protocol
the gateway allocates IP address using Dynamic Host Configuration (GTP) tunneling, and Proxy Binding Update (PBU) and Proxy
Protocol (DHCP) and the MN updates its Domain Name System Binding Acknowledgment (PBA) messages are exchanged between
(DNS) record using the IP address (e.g. prefix1). The DMM-GW S-GW and P-GW instead of Modify Bearear Request and Response
sends a DNS query to obtain the IP address that the MN has to messages. The adoption of PMIPv6 in the 3GPP EPC architecture is
use. It advertises then all the routers in the network that the specific shown at the Figure 7.
address is reachable via the corresponding DMM-GW. When han- In fact, Release 8 [32] describes the attachment of UE to EPC.
dover occurs, the new DMM-GW confirms that the MN has already The function of MAG is sending a PBU to the LMA. Thus, the P-GW
an IP address (DNS lookup), then using the BGP protocol, the pre- uses tunneling technique for the downlink packets of UE to avoid the
fix1 is announced in the routing system (BGP routing update). The problem of packet loss. The P-GW provides an IP address/prefixes in
IP address is kept unchanged during a handover. Thus, the mecha- the Proxy Binding Agent (PBA) to the UE. The P-GW uses GRE key,
nism relies on BGP UPDATE to announce the new route. The new and the S-GW also uses tunneling technique for the uplink packets of
DMM-GW will update the routers in the network so the traffic will UE. During a handover, a GRE tunnel is created between the LMA
be able to reach the new location of the MN. (P-GW) and the new MAG (S-GW), enabling packet communication
over the new path. The same IP addresses are used before and after
3.2 3GPP based DMM the handover.

In this section, we briefly introduce the existing approaches Local IP 3.3 SDN based DMM
Address/Selected IP Traffic Offload (LIPA/SIPTO) [28, 29] in 3GPP
to handle some DMM solutions to alleviate the load of the mobile SDN is a new network design [33] which attempts to build a network
core network. by separating it into two systems. The first system is the control
plane which provides performance and fault management. It typ-
3.2.1 LIPA/SIPTO: SIPTO enables an operator to offload cer- ically handles configuration management to the SDN devices and
tain types of traffic at a network node close to the UE’s point of understands the network topology. Loaded with those details, the

IET Research Journals, pp. 1–13

c The Institution of Engineering and Technology 2015 5

2018/07/30 17:04:37 IET Review Copy Only 6

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

(a) SIPTO solution (b) LIPA solution

Fig. 6: 3GPP based DMM

Recently, Nguyen et al. [40] present an SDN-based DMM (S-

DMM) solution provided as a service on top of the SDN controller.
As shown in Figure 9 the Home Subscriber System (HSS), Serv-
ing Gateway (S-GW), and Packet Data Network Gateway (P-GW)
are implemented in the SDN controller while the data plane is man-
aged by the basic OpenFlow Switches (OFSs). The DMM, the AAA
(Authentication, Authorization and Accounting) as well as the pol-
icy profile service are deployed as SDN applications on the top of the
controller. The OFS can detect the new attachment of a new MD by
means of discovery messages. It asks then the SDN controller about
the forwarding rules and other information related to MD through
Packet-in message. The S-DMM entity receives the message and
then updates its database to keep track the location and the MD’s
information. A Router Advertisement (RA) message is then created
and sent to the corresponding OFS via an OpenFlow protocol mes-
sage (Packet-out). To start communication with the corresponding
node, the OFS forwards in turn the packet to MD. Concerning the
Fig. 7: PMIPv6 application in the 3GPP EPC architecture SDN controller, it receives the first packet of the flow and updates
the forwarding rules based on the routing services. Those rules are
sent to the OFS to reroute all the flow to the destination.
In case of a handover between two OFSs, the same procedure will
be executed to update MD address, prefix and location but the RA
controller can process connection requests based on desired require- will be sent to the new OFS. The controller distributes the for-
ments such as QoS levels. The controller can also perform link warding rules also to the two switches. The flows will be rerouted
management between devices. The second system is the data plane between the two switches using two different modes. The first mode
which is responsible for forwarding traffic to the selected destina- is the Tunneling mode (ST-DMM), similar to the legacy DMM
tion. Otherwise, the control plane configures connection paths or approaches. The second is called Optimal mode (SO-DMM) in
flows into the data plane to the use of the control protocols [34]. which the controller calculates the optimal new route and distributes
The control protocol is used by the controller in the SDN to perform the forwarding tables to all the OFSs placed between the new OF
functions such as connection setup. When a host attempts to commu- switch and the corresponding node.
nicate with another host over SDN, the first packet involved with the According to the numerical results and the test-bed, the pro-
new flow is used to determine whether a decision can be made locally posed S-DMM solution shows its performance in terms of handover
by a switch, or if the switch needs to ask the controller what to do latency and end-to-end delay comparing with the legacy DMM
via a secure channel using the control protocol OpenFlow [35]. One approaches which make it a candidate for mobility management
important reason for the wide attention paid to SDN is that it is just solution in 5G networks.
a paradigm, instead of an ossified architecture. Indeed, it is rational
for researchers to adopt the main idea of SDN in many other fields
and not being limited in one fixed structure. Consequently, SDN has
been extended and applied to various scenarios, such as computer 4 Discussion
networks [36], optical networks, wireless scenarios [37, 38], etc.
The SDN-based DMM solution applied in D2D communication In this section, we summarize the given DMM technologies with
is shown in Figure 8 and studied by Yazici et al. [19] and Liu et al. a qualitative comparison between them and compared on different
[39]. Thus, a new entity called network controller is added which characteristics such as tunneling, scalability, handover latency or
is responsible of tracking the D2D mobile devices and forwarding signaling overhead as shown in Table 2.
rules to the DMM gateways (the access BSs) via the OpenFlow API. The ’DMM category’ is an essential criteria in mobility manage-
Hence, upon the attachment of both or one D2D mobile to a new BS, ment since we aim to avoid the centralized one. It includes fully or
the controller configures rules in each BS visited by the D2D pairs partially distributed mobility. ’D2D support’ is also an interesting
and thereby, the session continuity is maintained since there is no characteristic. Based on it, we choose the best technology that suits
change in IP address. This solution can be partially distributed, as the D2D features. ’Mobility anchors’ represents the base of a cen-
shown in Figure 8(a), or fully distributed as shown in Figure 8(b). tralized mobility. The ’Network-based’ characteristic refers to not

IET Research Journals, pp. 1–13

6 c The Institution of Engineering and Technology 2015

2018/07/30 17:04:37 IET Review Copy Only 7

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

(a) Partially distributed SDN architecture (b) Fully distributed SDN architecture

Fig. 8: SDN based DMM solution to support D2D communications

Table 2 Comparison between DMM solutions

DMM features PMIPv6-based Routing-based LIPA SDN-based
DMM category partially or fully fully partially partially
Routing optimization + + - -
Real-time application + - + +
Handover latency low high not considered low
Signaling overhead low high not considered low
Mobility anchors + - + +
Co-existence - + + +
Security considerations + depending on the routing protocol + -
Dynamicity + - - -
External mechanism + + + -
Tunneling - - - +
Packet Delay low low high high
IPv6 deployment + + + +
Gradual deployment easy difficult easy easy
Reliability controller gateway (router) L-GW DMM-GW
Additional cost no cost propagating routes no cost tunneling overhead
Packet loss rate low high low high
Scalability + - + +
Type of mobility management Network-based Network-based Network-based Network-based
D2D support + - + +

properties of the solution. We compare also the different technolo-

gies regarding few common characteristics: reliability, deployment,
dynamicity and packet delay.
Based on Table 2, the SDN based DMM satisfies almost all the
requirements and outperforms the routing-based solutions, PMIPv6-
based DMM and LIPA. The current PMIPv6-based mobility solution
face some obstacles to fit into a D2D communication and distributed
network architecture since they need IP tunnels when a handover
occurs and an overhead is added due to packet IPv6 encapsulation in
contrast to the routing based DMM and SDN.
Routing-based proposals would not be suitable for large networks
since the whole network must be informed of every MD’s movement
and it is feasible only in small networks with a small number of MDs.
Those solutions inherit the issues related to handover latency, sig-
naling overhead and scalability (storms of routing updates). Router
protocols (BGP) are also not designed to react immediately in order
Fig. 9: S-DMM architecture to reduce the ping-pong effects and messages’ flooding. By the way,
the signaling overhead can be very high during a handover if the
D2D pairs move to another attachment points (access routers) in dif-
ferent cells since the routing update must be forwarded to a large
involving MD in the mobility and handover processes. The ’Scalabil- part of the network. It also causes a large delay and HO latency due
ity’ indicates whether the proposed DMM solution is able to handle to the DNS mapping of the D2D pair’s.
an increasing number of mobile devices easily without affecting the

IET Research Journals, pp. 1–13

c The Institution of Engineering and Technology 2015 7

2018/07/30 17:04:37 IET Review Copy Only 8

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

On the other hand, even if LIPA reduces network congestion and IP Connectivity Access Network (IPâĂŞCAN) Session Modification
increases network revenues, it is not suitable for our case study messages.
since the Wi-Fi exists for long. In fact, this 3GPP solution is not
very useful for our days MDs as Wi-Fi already provides this at 5.1.1 Chen et al.: To analyze the Chen et al. handover delay,
greater speed and in a more straight forward manner. The sensors we use Figure 2(a) as a benchmark and based on [41]:
nodes or any other device type that can not use the Wi-Fi can use
LIPA but they should be grouped in a small area to create small THO P rep = 2 TSeN B7→T eN B + teN B (2)
cells. In addition, some additional software modifications in MDs TSeN B7→T eN B is the transmission delay between the target and
are required when they send a LIPA flag during the establishment serving eNBs, and teN B represents the processing delay of the eNB.
of the PDP context/default bearer to get a connection to the local
network. SIPTO and LIPA with LIMONET only provide localized THO Ex = 2 THIT + 2 TU E7→eN B (3)
mobility support which requires the deactivation and re-activation of
the network connections when MD is not moving locally. Thus, this THIT is the handover interruption time. The measurement of HIT
leads to performance degradation of the service continuity in case of is obtained by computing the time difference between the recep-
a D2D communication. tion of the RRC Connection Reconfiguration and the reception of the
The SDN is a promising paradigm and seems to be the most suit- completion acknowledging RRC Connection Reconfiguration Com-
able to solve the D2D management mobility issue since we have the plete. TU E7→eN B indicates the transmission delay between the D2D
ability to introduce many new functions related to D2D communi- terminal and the corresponding eNB.
cation due to its flexibility. Handover delay and packet loss rate are
reduced in SDN since the controllers update dynamically the routing THO Comp = 2 TeN B7→M M E + 2 TM M E7→P −GW +
tables in the routers and hence the time spent in exchanging mes- TIP.CAN + TSeN B7→T eN B +
sages is considerably decreased. The overhead is likewise low as tS−GW + tM M E + teN B
messages are exchanged only between two entities to which MDs (4)
are connected. Energy consumption also decreases since MDs are TIP.CAN is the processing delay of IP–CAN Session Modification.
not involved in the mobility management. Moreover, the SDN con- tM M E and tS−GW are the processing delay of the MME and the
troller does not need any IP tunnels and has the network oversight, S-GW respectively. TeN B7→M M E and TM M E7→P −GW are the
so routing optimization and security are guaranteed. However, the transmission delay between the eNB and the MME and between the
SDN can not coexist with other technologies due to the introduction MME and the P-GW.
of new network entities (controllers and switchers). Similar to the above analysis, based on Figure 2(b), we ana-
Finally, based on this qualitative analysis, we argue that the lyze the handover latency of the D2D half handover procedure. The
PMIPv6 and SDN-based behave quite similar in terms of perfor- only difference between the joint and half procedure resides in the
mance and can provide mobility in a timely manner. They are faster handover execution phase and can be formulated as:
to respond to the network change but they require specialized entities
to perform the needed functions. In order to justify this appropriate THO Ex = 2 THIT + 2 TU E7→eN B + T T TD (5)
technology for D2D communication in 5G networks, we propose to where T T TD is the time to trigger the D2D handover.
evaluate analytically the HO performance of the SDN-based mech-
anism and PMIPV6. Theses solutions are compared with the Chen 5.1.2 3GPP-PMIPv6: As described in [42, 43], based on Fig-
HO schemes in LTE networks [18]. ures 7 and 10, and during a handover completion process the target
eNB sends path switch request message to the MME. The MME
sends update bearer request message to serving gateway (S-GW).
5 Analytical Modeling Then serving gateway sends Proxy Binding Ipdate (PBU) message
to P-GW which in turn reply with a Proxy Binding Agent (PBA)
In this section, the performance of SDN-based DMM is evaluated message.
and compared with PMIPv6 and Chen handover mechanism [18]
based on analytical basis and under different parameters. Evaluation
criteria to be considered in this paper are the handover latency, the
packet loss, ping-pong effect and the handover failure probability
which are the most important performance metrics in the mobility
management protocols. The handover delay is defined as the time
interval between the time when D2D candidate can not receive con-
trol packets from the SeNB and the time when it receives the first
packet from the TeNB. The packet loss is defined as the number of
packets lost during the handover procedure.

5.1 Handover Delay

Fig. 10: PMIPv6 joint D2D handover completion signaling call flow
The handover delay is generally expressed as follows [41]:

Hence, the D2D joint handover latency for PMIPv6 can be

HD = THO P rep + THO Ex + THO Comp (1)
calculated as:

where THO P rep is the average time needed for the HO Preparation
phase which is performed without core network elements so that the THO Comp = 2 TeN B7→M M E + 2 TLU +
preparation messages are directly exchanged among the eNBs and TSeN B7→T eN B + TIP.CAN + (6)
MD. It includes the transmission of the Handover Request and Han- tS−GW + tM M E + teN B + tP −GW
dover Request Ack messages. where TLU represents the time taken by the PBU and PBA mes-
THO Ex is the average time needed for the HO execution phase. It sages, and it can be defined as:
is measured from sending the Radio Resource Control (RRC) Con-
nection Reconfiguration to MD to the reception of RRC Connection TLU = 2HS−GW 7→P −GW (Wp + Lα ) (7)
Reconfiguration Complete message by the TeNB.
THO Comp is the average time needed for the HO Completion HS−GW 7→P −GW is the number of hops between the S-GW and P-
phase. It consists of sending the Path Switching, Modify Bearer, and GW. The tP −GW is the service delay at the gateway once it receives

IET Research Journals, pp. 1–13

8 c The Institution of Engineering and Technology 2015

2018/07/30 17:04:37 IET Review Copy Only 9

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

a message. The gateway is hence modeled as a M/M/1 queue with 5.2 Packet loss
service rate µP −GW and arrival rate λP −GW [42]. The mean wait
time can then be formulated as During a handover, MD stops receiving the packets as soon as it
disconnects from the previous eNB, and resumes after receiving
1
µP −GW the packets from the new eNB. If any buffering mechanism is not
WP −GW = λP −GW
(8) supported, data packets destined to MD will be lost when it under-
1− µP −GW goes its handover. Hence the packet loss of PMIPv6-based and Chen
mechanisms can be calculated as described in [42]:
Wp is the mean wait time of the network nodes and can be calcu-
lated as eq.8 with with arrival rate equal to λp . The THO P rep and
P LP M IP v6 = λs E(S)HDP M IP v6
THO Ex are identical to those shown in Chen HO mechanism and (12)
P LChen = λs E(S)HDchen
calculated as shown in eq. 2 and 3.
For the D2D half handover procedure in PMIPv6, the HO prepa-
ration phase is calculated based on equation 2, the HO execution where the λs is the average session arrival rate and the E(S) is the
follows the equation 5 and the HO completion phase is formulated average session length [45]. The seamless SDN mechanism utilizes
as 6. the buffering scheme [43]. Let Bs denote the packet buffer size of
the OF switches and the controller. The number of lost packet during
5.1.3 SDN: The proposed SDN scheme consists of a central- handover of SDN is:
ized controller and OF Switches as shown in Figure 11. The S-GW
functionality is split into control and user planes (S-GW/U-C). The
P LSDN = λs E(S)HDSDN − Bs (13)
S-GW/U is serving as an OF switch. The use of GTP or PMIPv6 tun-
nels for IP mobility is eliminated. The components of the proposed
architecture are described in section 3.3. 5.3 Handover Failure Probability
We propose here a D2D joint handover mechanism. The HO
preparation and execution phases are identical to the Chen HO mech- We define the Handover Failure Probability (PH F ) as the probability
anism. The SDN controller and the OpenFlow switch (OFS) take part that the cell residence time (Tc ) is less than the handover latency
only at the handover completion phase. We present in Figure 11) an [46]. Its general expression is defined as:
overview of a D2D joint handover completion process.
PHF(.) = P (Tc ≤ HD(.) )
RHD(.)
= 0 µc e−µc x dx (14)
= 1 − e−µc HD(.)

Note that (.) is used as a mechanism indicator. (Tc ) is a random vari-

able which follows an exponential distribution with mean rate µc .
According to [47], µc is expressed as:

E(v)Pe
µc = (15)
πA

Fig. 11: SDN joint D2D handover completion signaling call flow E(v) is the average of the speed v of the D2D MDs. Pe is the length
of the cell’s perimeter and A is the cell’s area.
For the D2D half handover, the handover failure probability follows
When the controller receives Path Switch Request message, it the half handover latency for each technology.
sends an OF Modify Flow Entry message to update the correspond-
ing entry at the OF switch [44]. The packets destined to the controller
are not associated with any flow at any OF switch of the network. 5.4 Ping-Pong effect
Consequently, they will be encapsulated in an OF Packet IN message
and directly sent to the SDN controller via the OpenFlow interface. The Ping-Pong effect is defined as the unnecessary handover to the
The S-GW/U sends a Path Switch Request Ack message to the TeNB neighbouring eNodeB while MD is returning to the original BS after
to finish the HO process. Packets forwarded by the source eNB are a very short span of time. The time tag (Tg ) is defined as the time-
buffered by the TeNB during the HO interruption time (HIT). period that has elapsed from the last successful handover. As time
The handover completion latency is calculated as follows: tag increases the ping-pong effect reduces.
The Ping-Pong probability is generally expressed as follows [47]:
THO Comp = TSeN B7→T eN B + 2 TeN B7→OF S +
2 TOF S7→Con + tOF S + teN B + tCon 1 2
PP ing−P ong = P (Tg ≤ HD(.) + HD(.) ) (16)
(9)
where tOF S , tCon are the processing delay in the OF switch and the
SDN controller respectively and can be expressed as shown in eq.8. HD1(.) and HD2(.) are respectively the first and second consecutive
The communication delay between the eNB and the SDN controller handover latencies. Supposing that HD1(.) =HD2(.) , so the Ping-Pong
is encompassed by the TeN B7→OF S and TOF S7→Con . The commu- effect is:
nication delay between the OF switch (MAG) and the controller is
encompassed by the TOF and defined based in Figure 11. PP ing−P ong(.) = P (Tg ≤ 2HD(.) )
R2HD(.)
TeN B7→OF S = 2 HeN B7→OF S (Wp + Lα ) (10) = 0 µg e−µg x dx (17)
= 1 − e−2µg HD(.)
TOF S7→Con = 2 HOF S7→Con (Wp + Lα ) (11)
where (Tg ) is assumed to be exponentially distributed with mean rate
HeN B7→OF S and HOF S7→Con are the numbers of hops between
µg . µg is expressed as:
the eNB and the OF switch and between the OF switch and the SDN
controller respectively. 0
For the D2D half handover preparation, execution and completion E(v)Pe
phases, we follow the equations 2, 5 and 9 respectively. µg = 0 (18)
πA

IET Research Journals, pp. 1–13

c The Institution of Engineering and Technology 2015 9

2018/07/30 17:04:37 IET Review Copy Only 10

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

Table 3 Performance Evaluation Parameters

Parameters Description Values
HS−GW 7→P −GW ,HeN B7→OF S , HOF S7→Con Number of hops between S-GW, P-GW and eNB, OFS and Controller respectively 1-19
λP −GW , λOF S , λCon Packet arrival rate at P-GW, OFS and Controller, respectively 1-9 Mbps
λs Average session arrival rate 0.7 s
E(S) Average session length 20
Lβ Wireless link propagation delay 15 ms
Lα Wired link propagation delay 0.0075 ms
E(v) Average velocity 2 m/s
R Cell radius 100 m
λp Packet arrival rate at the network nodes 1-9 Mbps
µp Link service rate 10 Mbps
teN B Processing delay of the eNB 0.5-5 ms
tS−GW Processing delay of the S-GW 4 ms
tM M E Processing delay of the MME 4 ms
tOF S Processing delay of the OF switch 1 ms
tCon Processing delay of the SDN controller 1 ms
T T TD Time To Trigger the D2D HO 2 ms
THIT Handover interruption time 17 ms
TSeN B7→T eN B Transmission delay between a SeNB and a TeNB 5 ms
TU E7→eN B Transmission delay between a UE and an eNB 2 ms
TeN B7→M M E Transmission delay between an eNB and a MME 1 ms
TM M E7→P −GW Transmission delay between a MME and a P-GW 6 ms
TIP.CAN Processing delay of IPâĂŞCAN Session Modification 18 ms

Fig. 12: Intersection zone

0 0
where Pe is the length of the intersection zone’s perimeter and A its
area. Based on Fig. 12, we can deduce that:
0 π
Pe = 2R (19)
3 Fig. 13: Handover delay Comparison

0 π π
A = R2 ( − sin( )) (20)
3 3
For the D2D half handover, the ping-pong effect follows the half
handover latency for each technology.

6 Performance Evaluation
It is worth mentioning that an efficient handover scheme in cellular
networks able to manage a D2D mobility is achieved with a low han-
dover latency and low packet loss. In the following, we illustrate our
simulation results. The values of parameters presented in the Table 3
are used in the evaluation of the analytical model [41, 42, 47].
The handover latency comparison between the SDN solution,
PMIPv6 and Chen mechanism is shown in the Figure 13, where the
processing delay of the eNB (teNB) is considered as the variable
factor. As the processing delay of the eNB, we see that the handover
performance of SDN scheme improves up to 30%. The SDN-based
handover is almost faster than the PMIPv6 protocol and Chen mech-
anism. The reason is that the PMIPv6-based and Chen mechanisms
operate in a conceptually similar way. Fig. 14: CDF of the Handover delay measurements
Figure 15 shows the variation of the packet loss as a function of
the processing delay teN B . Compared to PMIPv6 and Chen’s pro-
posed handover, the SDN prevents the packet loss as a consequence With lower processing delay, the eNodeB has the possibility to
of adopting a buffering mechanism. quickly finish the handover process and hence the D2D terminals

IET Research Journals, pp. 1–13

10 c The Institution of Engineering and Technology 2015

2018/07/30 17:04:37 IET Review Copy Only 11

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

Fig. 15: Packet loss Comparison Fig. 17: Handover Failure Probability Comparison

Fig. 16: Ping-Pong effect Comparison

have more time to complete their handovers before switching to the

target eNodeB. The results show in Figure 16 and Figure 17 that
SDN-based solution outperforms PMIPv6-based and Chen mecha-
nisms by reducing the handover failure probability as well as the Fig. 18: SDN relative D2D Handover performance gain
ping-pong effect. This is due to less handover latency in SDN than
in the other processes. Thus, we can argue that SDN maintains better
QoS for D2D handover sessions. and [18] suffer from packet losses due to lack of any buffering tech-
The relative handover performance gain of the SDN over the nique. And thirdly, they are caused by the SDN controller which
PMIPv6 and Chen mechanisms is calculated as [42, 48]: updates the flow tables in the OFSs. The time spent in exchang-
ing messages is considerably decreased and the overhead is likewise
HDP M IP v6 low as the communication become between only network entities in
P GH = (21) which the D2D candidates are connected.
HDSDN
The results in the Figure 19 show that the PMIPv6 handover
latency exponentially increase with the increase in network utiliza-
HDChen
P G0H = (22) tion. The SDN process follows the same trend as the PMIPv6 but
HDSDN has 35% less handover latency when ρp ≤ 0.4. Hence, the pro-
For instance, if P GH is larger than 1.0, it means that the SDN posed SDN mechanism has a distinguishing characteristic which
joint D2D handover process outperforms the PMIPv6 one. First, is the capability to sustain under the high network traffic load and
we observe the variation of PGH (PG0H ) with different values of communication delays, where the PMIPv6 performance degrades.
HM AG7→LM A and network utilization. As shown in Figure 18,
the SDN handover process always outperforms the PMIPv6. As
HM AG7→LM A is increased, PGH decreased marginally, whereas 7 Conclusion and Future Work
PGH is out of proportion to the increase of ρp . It is seen that
HM AG7→LM A is sensitive to PGH , whereas ρp does not much In this paper, we have surveyed different works for D2D handover
affect. management. It has shown that advanced architectures are needed to
These phenomenons are caused by firstly the SDN process relying perform better mobility management making for user satisfaction as
on few messages compared to the PMIPv6 one, secondly by the fast well as for the efficient use of network resources. We have focused
data packet forwarding of the SDN D2D handover process. In fact, essentially on DMM as a promising approach in 5G cellular net-
the SDN prevents packet losses during handovers, whereas PMIPv6 works. We have then shown that the SDN-based DMM solution is

IET Research Journals, pp. 1–13

c The Institution of Engineering and Technology 2015 11

2018/07/30 17:04:37 IET Review Copy Only 12

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

IP Internet Protocol
IPv6 Internet Protocol version 6
L-GW Local-Gateway
LIMONET LIPA MObility and SIPTO at the Local Network
LIPA Local IP Access
LTE-A Long Term Evolution-Advanced
LMA Local Mobility Anchor
LMD Localized Mobility Domain
MAG Mobile Access Gateway
MD Mobile Device
MIPv6 Mobile IP
MME Mobility Management Entity
MN Mobile Node
NFC Near Field Communication
NFV Network Function Virtualization
OFS OpenFlow Switches
P-GW Packet Data Network GateWay
PBA Proxy Binding Acknowledgement
PBU Proxy Binding Update
Fig. 19: Handover latency comparison with increasing utilization of PCRF Policy and Charging Rules Function
network elements PDP Packet Data Protocol
PMIP6 Proxy Mobile IPv6
Prose Proximity Services
the suitable candidate used to manage D2D mobility and achieves QoE Quality of Experience
30% improvement over the standard PMIPv6 in terms of handover QoS Quality of Service
latency and 100% improvement in terms of packet loss. RAN Radio Access Network
Following this line, we propose, as future work, a sophisticated RA Router Advertisement
SDN architecture able to enhance and manage the D2D mobility in S-GW Serving GateWay
5G cellular networks. This work intends to answer D2D handover SDN Software Defined Networking
challenges as flexibility, efficiency, seamlessness, automation and SIPTO Selected IP Traffic Offload
performance optimization. TTTD Time To Trigger of D2D HO
UE User Equipment
UMTS Universal Mobile Telecommunications System
8 Abbreviations Wi-Fi Wireless Fidelity
WIMAX Worldwide Interoperability for Microwave Access
5G 5th Generation WLAN Wireless Local Area Network
3GPP 3rd Generation Partnership Project
AAA Authentication, Authorization and Accounting
API Application Programming Interface
AR Access Router
BGP Border Gateway Protocol 9 References
BS Base Station
1 Jeffrey G Andrews, Stefano Buzzi, Wan Choi, Stephen V Hanly, Angel Lozano,
C-Plane Control-Plane Anthony CK Soong, and Jianzhong Charlie Zhang. What will 5g be? IEEE Journal
C-RAN Cloud Radio Access Network on Selected Areas in Communications, 32(6):1065–1082, 2014.
CMAAS Connectivity Management As A Service 2 Michel Mouly, Marie-Bernadette Pautet, and Thomas Foreword By-Haug. The
CMD Control Mobility Database GSM system for mobile communications. Telecom publishing, 1992.
3 Stefania Sesia, Issam Toufik, and Matthew Baker. LTE-the UMTS long term
CN Core Network evolution. Wiley Online Library, 2015.
CN Corespondent Node 4 Jeffrey G. Andrews, Arunabha Ghosh, and Rias Muhamed. Fundamentals of
CoA Care of Addresses WiMAX: Understanding Broadband Wireless Networking. Pearson Education,
D2D Device-to-Device 2007.
5 Edney and William A. Arbaugh. Real 802.11 Security: Wi-Fi Protected Access and
D2I Device-to-Infrastructure 802.11 i. Addison-Wesley Longman Publishing Co., Inc., 2003.
DHCP Dynamic Host Configuration Protocol 6 Woon Hau Chin, Zhong Fan, and Russell Haines. Emerging technologies and
DMM Distributed Mobility Management research challenges for 5g wireless networks. IEEE Wireless Communications,
DNS Domain Name System 21(2):106–112, 2014.
7 Jiajia Liu, Nei Kato, Jianfeng Ma, and Naoto Kadowaki. Device-to-device com-
E2E End-to-End munication in lte-advanced networks: a survey. IEEE Communications Surveys &
EPC Evolved Packet Core Tutorials, 17(4):1923–1940, 2014.
GPRS General Packet Radio Service 8 Lei Lei, Zhangdui Zhong, Chuang Lin, and Xuemin Shen. Operator con-
GRE Generic Routing Encapsulation trolled device-to-device communications in lte-advanced networks. IEEE Wireless
Communications, 19(3):96, 2012.
GSM Global System for Mobile communications 9 Arash Asadi, Qing Wang, and Vincenzo Mancuso. A survey on device-to-device
GTP GPRS Tunneling Protocol communication in cellular networks. IEEE Communications Surveys & Tutorials,
GW GateWay 16(4):1801–1819, 2014.
HA Home Agent 10 Daniel Camps-Mur, Andres Garcia-Saavedra, and Pablo Serrano. Device-to-device
communications with wi-fi direct: overview and experimentation. IEEE wireless
HIT Handover Interruption Time communications, 20(3):96–104, 2013.
HeNB Home eNodeB 11 Arash Asadi and Vincenzo Mancuso. Wifi direct and lte d2d in action. In Wireless
HetNet Heterogeneous Networks Days (WD), 2013 IFIP, pages 1–8. IEEE, 2013.
HO HandOver 12 Shahid Mumtaz, Kazi Mohammed Saidul Huq, and Jonathan Rodriguez. Direct
mobile-to-mobile communication: Paradigm for 5g. IEEE Wireless Communica-
HSS Home Subscriber System tions, 21(5):14–23, 2014.
IETF Internet Engineering Task Force 13 Mohsen Nader Tehrani, Murat Uysal, and Halim Yanikomeroglu. Device-to-
device communication in 5g cellular networks: challenges, solutions, and future
directions. IEEE Communications Magazine, 52(5):86–92, 2014.
14 Lingyang Song, Dusit Niyato, Zhu Han, and Ekram Hossain. Game-theoretic
resource allocation methods for device-to-device communication. IEEE Wireless
Communications, 21(3):136–144, 2014.

IET Research Journals, pp. 1–13

12 c The Institution of Engineering and Technology 2015

2018/07/30 17:04:37 IET Review Copy Only 13

ReView by River Valley This
Technologies
article has been accepted for publication in a future issue of this journal, but has not been fully IET Communications
edited.
Content may change prior to final publication in an issue of the journal. To cite the paper please use the doi provided on the Digital Library page.

15 Qin Wang, Wei Wang, Shi Jin, Hongbo Zhu, and Nai Tong Zhang. Mode selection implementation in a 5g sdn-based mobile network architecture. In Personal,
for d2d communication underlaying a cellular network with shared relays. In Wire- Indoor, and Mobile Radio Communications (PIMRC), 2016 IEEE 27th Annual
less Communications and Signal Processing (WCSP), 2014 Sixth International International Symposium on, pages 1–6. IEEE, 2016.
Conference on, pages 1–6, China, 2014. IEEE. 45 Jong-Hyouk Lee, Zhiwei Yan, and Ilsun You. Enhancing qos of mobile devices by
16 Meriem Kassar, Brigitte Kervella, and Guy Pujolle. An overview of verti- a new handover process in pmipv6 networks. Wireless Personal Communications,
cal handover decision strategies in heterogeneous wireless networks. Computer 61(4):591–602, 2011.
Communications, 31(10):2607–2620, 2008. 46 Hassan Ali-Ahmad, Meryem Ouzzif, Philippe Bertin, and Xavier Lagrange. Per-
17 Osman NC Yilmaz, Zexian Li, Kimmo Valkealahti, Mikko A Uusitalo, Martti formance analysis on network-based distributed mobility management. Wireless
Moisio, Petteri Lundén, and Carl Wijting. Smart mobility management for d2d personal communications, 74(4):1245–1263, 2014.
communications in 5g networks. In Wireless Communications and Networking 47 Kaouthar Ouali, Meriem Kassar, Thi Mai Trang Nguyen, Kaouthar Sethom, and
Conference Workshops (WCNCW), 2014 IEEE, pages 219–223. IEEE, 2014. Brigitte Kervella. Modeling d2d handover management in 5g cellular networks.
18 Ho-Yuan Chen, Mei-Ju Shih, and Hung-Yu Wei. Handover mechanism for device- In Wireless Communications and Mobile Computing Conference (IWCMC), 2017
to-device communication. In Standards for Communications and Networking 13th International, pages 196–201. IEEE, 2017.
(CSCN), 2015 IEEE Conference on, pages 72–77. IEEE, 2015. 48 Jong-Hyouk Lee, Thierry Ernst, and Naveen Chilamkurti. Performance analysis
19 Volkan Yazıcı, Ulas C Kozat, and M Oguz Sunay. A new control plane for 5g of pmipv6-based network mobility for intelligent transportation systems. IEEE
network architecture with a case study on unified handoff, mobility, and routing Transactions on vehicular technology, 61(1):74–85, 2012.
management. IEEE Communications Magazine, 52(11):76–85, 2014.
20 Antonino Orsino, Giuseppe Araniti, Li Wang, and Antonio Iera. Enhanced c-ran
architecture supporting sdn and nfv functionalities for d2d communications.
21 3GPP. R2-140363: "d2d mobility scenarios". Technical report, 2014.
22 3GPP. R1-131216: "mobility requirements for general d2d scenarios". Technical
report.
23 Fabio Giust, Carlos J Bernardos, and Antonio De La Oliva. Analytic evaluation and
experimental validation of a network-based ipv6 distributed mobility management
solution. IEEE Transactions on Mobile Computing, 13(11):2484–2497, 2014.
24 H Chan, D Liu, P Seite, H Yokota, and J Korhonen. Rfc 7333:
Requirements for distributed mobility management. IETF. https://www.rfc-
editor.org/rfc/rfc7333.txt, 2014.
25 Dong-Hoon Shin, Danny Moses, Muthaiah Venkatachalam, and Saurabh Bagchi.
Distributed mobility management for efficient video delivery over all-ip mobile
networks: Competing approaches. IEEE network, 27(2):28–33, 2013.
26 Morteza Karimzadeh, Luca Valtulina, and Georgios Karagiannis. Applying
sdn/openflow in virtualized lte to support distributed mobility management (dmm).
2014.
27 Fabio Giust, Luca Cominardi, and Carlos J Bernardos. Distributed mobility man-
agement for future 5g networks: overview and analysis of existing approaches.
IEEE Communications Magazine, 53(1):142–149, 2015.
28 3GPP. Tr 23.401: General packet radio service (gprs) enhancements for evolved
universal terrestrial radio access network (e-utran) access. Technical report,
September 2011.
29 3GPP. Tr 23.829: Local ip access and selected ip traffic offload (lipa-sipto) (release
10). Technical report, October 2011.
30 3GPP. Tr 23.859: Lipa mobility and sipto at the local network. Technical report,
April 2013.
31 Julien Laganier, Takeshi Higuchi, and Katsutoshi Nishida. Mobility management
for all-ip core network. NTT Docomo Technical Journal, 11(3):34–39, 2009.
32 3GPP. Ts23.402: "architecture enhancements for non-3gpp accesses". Technical
report, January 2009.
33 Software-defined networking: The new norm for networks. [online]
Available: https://www.opennetworking.org/images/stories/down-
loads/sdn-resources/white-oaoers/wo-sdn-newnorm.odf. accessed 10,
2014.
34 Bruno Astuto A Nunes, Marc Mendonca, Xuan-Nam Nguyen, Katia Obraczka,
and Thierry Turletti. A survey of software-defined networking: Past, present, and
future of programmable networks. IEEE Communications Surveys & Tutorials,
16(3):1617–1634, 2014.
35 Open Networking Foundation (ONF). Openflow configuration
and management protocol of-config 1.1. [Online] Available:
http://www.opennetworking.org/images/stories/downloads/of-config/of-config-
1.1.pdf. accessed 2012.
36 Intidhar Bedhief, Meriem Kassar, and Taoufik Aguili. Sdn-based architecture chal-
lenging the iot heterogeneity. In Smart Cloud Networks & Systems (SCNS), pages
1–3. IEEE, 2016.
37 Tewelde Degefa Assefa, Rashedul Hoque, Elias Tragos, and Xenofontas Dim-
itropoulos. Sdn-based local mobility management with x2-interface in femtocell
networks. In Computer Aided Modeling and Design of Communication Links
and Networks (CAMAD), 2017 IEEE 22nd International Workshop on, pages 1–6.
IEEE, 2017.
38 Songlin Sun, Michel Kadoch, Liang Gong, and Bo Rong. Integrating network
function virtualization with sdr and sdn for 4g/5g networks. IEEE Network,
29(3):54–59, 2015.
39 Jiajia Liu, Shangwei Zhang, Nei Kato, Hirotaka Ujikawa, and Kenichi Suzuki.
Device-to-device communications for enhancing quality of experience in software
defined multi-tier lte-a networks. IEEE Network, 29(4):46–52, 2015.
40 Tien-Thinh Nguyen, Christian Bonnet, and Jérôme Harri. Sdn-based distributed
mobility management for 5g networks. In Wireless Communications and Network-
ing Conference (WCNC), 2016 IEEE, pages 1–7, Doha, Qatar, 2016. IEEE.
41 Donghyuk Han, Sungjin Shin, Hyoungjun Cho, Jong-moon Chung, Dongseok Ok,
and Iksoon Hwang. Measurement and stochastic modeling of handover delay
and interruption time of smartphone real-time applications on lte networks. IEEE
Communications Magazine, 53(3):173–181, 2015.
42 Syed M Raza, Dongsoo S Kim, Dongryeol Shin, and Hyunseung Choo. Lever-
aging proxy mobile ipv6 with sdn. Journal of Communications and Networks,
18(3):460–475, 2016.
43 Yuh-Shyan Chen, Tong-Ying Juang, and Yao-Tsu Lin. A secure relay-assisted
handover protocol for proxy mobile ipv6 in 3gpp lte systems. Wireless Personal
Communications, 61(4):629–656, 2011.
44 Jonathan Prados-Garzon, Oscar Adamuz-Hinojosa, Pablo Ameigeiras, Juan J
Ramos-Munoz, Pilar Andres-Maldonado, and Juan M Lopez-Soler. Handover

IET Research Journals, pp. 1–13

c The Institution of Engineering and Technology 2015 13

2018/07/30 17:04:37 IET Review Copy Only 14