(Information Science Institute,State Key Laboratory of Integrated Service Networks,Xidian University,Xi’an 710071,China)

Software⁃Defined Cellular Mobile Network Solutions

Jiandong Li,Peng Liu,and Hongyan Li

(Information Science Institute,State Key Laboratory of Integrated Service Networks,Xidian University,Xi’an 710071,China)

The emergency relating to software⁃defined networking(SDN),especially in terms of the prototype associated with OpenFlow,pro⁃vides new possibilities for innovating on network design.Researchers have started to extend SDN to cellular networks.Such new programmable architecture is beneficial to the evolution of mobile networks and allows operators to provide better services.The typical cellular network comprises radio access network(RAN)and core network(CN);hence,the technique roadmap diverges in two ways.In this paper,we investigate SoftRAN,the latest SDN solution for RAN,and SoftCell and MobileFlow,the latest solu⁃tions for CN.We also define a series of control functions for CROWD.Unlike in the other literature,we emphasize only software⁃defined cellular network solutions and specifications in order to provide possible research directions.

SDN;cellular network;radio access network;core network;OpenFlow

1 Introduction

Software⁃defined networking(SDN)is a compelling technique that has spurred innovative experiments and the evolution of computer networks.It has gained more and more recognition in both academia and industry.SDN gives network designers greater flexibility to separate the control plane,which decides how traffic is han⁃dled,from the data plane,which forwards traffic according to decisions made by the control plane[1].SDN architecture makes the network programmable and facilitates the designing of new protocols.The control plane communicates with the da⁃ta plane via a well⁃defined API in order to direct packet⁃for⁃warding.A good example of such an API is OpenFlow[2], which was developed at Stanford University and is a milestone in SDN.Before OpenFlow,there was a dynamic tension be⁃tween those who envisioned a fully programmable network and those who envisioned a more pragmatic network for practical applications.

Researchers started to apply SDN in other fields,such as wireless networking.Both OpenRoads[3]and OpenFlow Wire⁃less[4]were developed at Stanford University.The core idea of these platforms is to make wireless networks open by flattening out some vertical wireless networking techniques.A program⁃mable data plane is created so that subscribers experience seamless handover in a heterogeneous network.However,an open policy does not take into account the specifications of dif⁃ferent access techniques,i.e.,differences between WLANs and cellular networks,and does not take into account explicit com⁃mercialrequirements.Gradually,two areasofresearch emerged:software⁃defined WLAN and software⁃defined cellu⁃lar networks.A typical WLAN solution is Odin[5].

Cellular networks,the main topic of this paper,initially had a relatively complex structure.They are used to guarantee QoS and generate revenue for operators.Before flowing to the Inter⁃net,traffic first passes through radio access networks(RANs) and then through core networks(CNs).A RAN comprising us⁃ers and base stations is generally responsible for radio⁃related services,such as radio resource mapping and interference man⁃agement.A CN lies between the base station and Internet and provides packet⁃or traffic⁃related carrier service,such as traf⁃fic classification and authorization.To facilitate programmabili⁃ty in cellular networks,it is necessary to address the unique⁃ness of the network[6].

SoftRAN[7],as its name suggests,is an SDN prototype de⁃signed to address the challenges in a RAN.Typically,interfer⁃ence management in an LTE mobile network is distributed by coordinating between cells.By separating the control plane from data plane and building a central controller,interference can be effectively migrated between cells.By comparison,the evolution of CN has been more complicated and challenging because fine⁃grained packet handling can also cause new prob⁃lems with scalability.SoftCell[8]and MobileFlow[9]are two recent attempts to counter this.

In this paper,we articulate three leading SDN solutions for a cellular network:SoftRAN,SoftCell and MobileFlow.These have been developed over the past two years and have guided new research and applications.We also introduce connectivity management for energy⁃optimized wireless dense networks (CROWD)[10],which is a collaborative project funded by the

European Commission under the Seventh Framework Pro⁃gramme(FP7).CROWD has a series of control functions that can be mapped into the actions in physical networks.We also discuss challenges related to the evolution of software⁃defined cellular networks and discuss our recent work.

2 Background

2.1 Software-Defined Networking and OpenFlow

SDN can be viewed in two ways:1)separation of the control plane from data plane and use of a single software⁃control pro⁃gram to control multiple data planes,and 2)abstraction of net⁃work control in terms of forwarding,specification,and distribu⁃tion[11].SDN enables flat,flexible data planes with high⁃level control planes.

The SDN layers are shown in Fig.1.The north part of the ar⁃chitecture refers to the part above the controller and includes the policy layer and application layer.The south part of the ar⁃chitecture refers to the programmable switches(e.g.,Open⁃Flow).The controller is the brain of the SDN,which obtains in⁃formation about global resources and network state from south interfaces and makes abstract and global decisions.The con⁃troller can be taken as a network operator system,and alterna⁃tives include NOX[12]and Floodlight[13].

One of the best⁃known south⁃part interfaces is OpenFlow switch,whose protocol has been standardized by the Open Net⁃working Foundation(ONF).OpenFlow is supported by many vendors,including HP,NEC and IBM,and associated switches are available on the market.By separating the control plane from the data plane,OpenFlow provides new possibilities for innovation.The main components of OpenFlow are the flow ta⁃ble and security channel(Fig.1).The flow table monitors and forwards packets.If the property of the packet matches exist⁃ing flow entries in the flow table,actions for that entry are per⁃formed on the packet.If there is no match,the switch delivers the packet to the security channel and communicates with the controller through the OpenFlow protocol.Finally,the control⁃ler determines how to process the packet and updates the entry in the flow table.

2.2 Architecture of an LTE Cellular Network

A typical mobile network comprises two main parts:RAN and CN.The RAN is responsible for radio⁃related functions,such as scheduling,radio resource and interference management,coding,and multiple antenna schemes.The CN is responsible for authen⁃tication,charging,and establishing end⁃to⁃end con⁃nections[14].Isolating these functions from the RAN is beneficial to integrating a CN with multiple RANs. Mobile terminals directly connect to the RAN(Fig. 2,right).A flat architecture with only one node,an eNodeB,is used in LTE.The eNodeB is a virtual eNode implemented as a site or base station.The eNodeB con⁃nects to MME over the control plane(Fig.2,solid line)and S⁃GW over user plane(Fig.2,dashed line).The eNodeB and MME communicate with each other via the X2 interface in or⁃der to manage intercell radio resources and coordinate inter⁃cell interference.

In CN,the mobility⁃management entity(MME)is the control⁃plane node.Its functions include connecting and releasing bearers to and from a terminal.The serving gateway(S⁃GW)is the user⁃plane node that connects the CN to the RAN.The S⁃GW acts as a mobility anchor for mobile management and pro⁃vides statistics for charging.The packet data network gateway (P⁃GW)is the edge node that connects the CN to the Internet. The P⁃GW allocates an IP address for a particular terminal.

3 Solutions of Soft⁃Defined Cellular Networks

3.1 RAN Solutions

The RAN ensures limited resources are used effectively in radio⁃related functions.The RAN allocates resources and man⁃ages interference,handover,and load⁃balancing.Currently, the control plane in a RAN is distributed and coordinates inter⁃cell interference through message exchange over X2 interface. This is not optimal partly because distributed coordination al⁃gorithms generally require iterative and periodic updates of ra⁃dio resource allocation decisions,and this is hard to get right at scale[7].SoftRAN boosts the utility by separating the con⁃trol plane from base stations and forming a high⁃level central controller.

SoftRAN conforms more to the second SDN concept men⁃tioned in section 2.1:abstraction.Base stations in a geographi⁃cally nearby area are abstracted as a big virtual station that has a global view of underlying base stations and controls their be⁃havior.SoftRAN defines resources in terms of time,frequency, and base stations.The central controller in the big station de⁃termines what spectrum and transmit power will be used at the sites of underlying base stations.The control signaling is de⁃

fined by the API and exchange between controller and under⁃ling base stations via the backhaul.

▲Figure 2.CN architecture(left)and RAN architecture(right).

There are latency problems in the backhaul,and wireless channel conditions may vary rapidly.To address these chal⁃lenges,SoftRAN also defines a local controller within base sta⁃tions.This local controller is responsible for decisions that do not affect neighboring sites.Following this principle,some functions are separated and specified;specifically,handovers and downlink transmit power in each channel are arranged by the controller because these have implications for nearby sites. The allocation of a resource block,i.e.,a minimum assignable resource unit in LTE,is noticeable.In the downlink,resource blocks can be allocated within base stations because the trans⁃mit power has been specified.Downlink resource block alloca⁃tion is mainly used because it is adaptable to channel varia⁃tions;however,uplink transmit power is controlled by users in order to counter path loss,and the central controller has no knowledge of the uplink transmit power.In the uplink,re⁃source block allocation dominates in the central controller.

We present the architecture of SoftRAN in Fig.3.A special component is RAN information base(RIB)containing global network state.It includes a weighted interference graph where the weight stands for the average channel conditions between the nodes.Flow Records stores flow⁃related information such as buffer state.Preferences indicate operator’s appetite for dif⁃ferent flows.

3.2 CN Solutions

In mobile communications,CN is unique in that it supports fine⁃grained service.Such support depends heavily on custom⁃ized policies based on a wide variety of subscriber attributes and application classes[8].These data services are usually pro⁃vided by a P⁃GW,which integrate network functions such as content filtering,traffic optimization,firewalls,and lawful inter⁃ception[15].However,combining all these functions on the da⁃ta plane in a P⁃GW may make a network inefficient,rigid,and complex.First,all traffic will be forwarded through P⁃GWs re⁃gardless of whether it is device⁃to⁃device traffic with latency re⁃quirements or video traffic with service⁃rate requirements. This increases delay and congestion.Second,if an operator wants to cancel an unneeded function in a P⁃GW,they have to replace the P⁃GWs.Finally,operators cannot adopt equipment from different vendors.

SDN is a good solution for abstracting the con⁃trol plane.Evolving cellular networks towards SDN presents new challenges,such as inade⁃quate scalability.SoftCell and MobileFlow help overcome such challenges.

3.2.1 SoftCell

The central idea of SoftCell is to decentralize the functions of a P⁃GW,offloading them to a se⁃rious of switches and middle⁃boxes,both of which have particular functions and application requirements and are controlled by a central controller.In keeping with the rationale of devolved responsibility,components between a base station and the Internet become low⁃cost,general equip⁃ment.The core network of SoftCell merely comprises middle⁃boxes,which act as transcoders,web caches or firewalls;ac⁃cess switches,which classify fine⁃grained packets from users; core switches,which include gateway switches that connect to the Internet and forward packets at high speed;and the control⁃ler,which makes global decisions[8].Thus,this architecture (Fig.4)is flatter and cheaper than a traditional LTE CN archi⁃tecture.In the SoftCell architecture,there are no special P⁃GWs or S⁃GWs(Fig.2,left),and the control⁃plane prototype is implemented on top of Floodlight[12].

To enable this architecture and fine⁃grained policy,the au⁃thors of[8]define service policy on top of the controller.This concept is similar to the policy layer in Fig.1.A service policy designates which traffic(in predicate)should be handled in what way(in action).In Table 1,the first clause indicates that video traffic deriving from a user with a gold billing plan must first pass through a firewall before going to the transcoder.The second clause indicates that M2M signaling should be given high priority when passing through a firewall in order to ensure low latency.

▲Figure 3.SoftRAN architecture.

▲Figure 4.SoftCell architecture.

Because there are too many different combinations of ser⁃vice requirements for different classes of traffic(e.g.,QoS classes,device types,etc.),a data explosion may occur in the flow table.To resolve this issue,SoftCell leverages the aggre⁃ gating multidimensional information as well as traditional loca⁃tion⁃and tag⁃based routing.Another issue to be addressed is the asymmetry of the architecture,i.e.,A CN usually connects to hundreds of or even thousands of base stations while only de⁃ploying a few gateway switches that connect to the Internet. Moreover,base stations only deal with the packets of a limited number of active users whereas gateway switches process the packets from the Internet,and the volume of data is usually large.Similarly,if we classify packets on both edges,the laten⁃cy will be large on the side closest to the Internet.SoftCell ad⁃dresses this issue by a concept called“smart access edge, dumb gateway edge.”In other words,SoftCell classifies pack⁃ets in access switches when a user invokes a flow and embeds associated messages in the head of the packet.Finally,the au⁃thors[8]highlight the problem of dynamics,which a special problem in wireless networks.This problem is solved by de⁃ploying a local agent or local controller to form a hierarchical control structure.Actually,this problem can be solved by intro⁃ducing SoftRAN in a complementary way.

3.2.2 MobileFlow

MobileFlow emphases evolution and provides a blueprint for software⁃defined mobile networks.It also separates the control plane from the data plane,which is a concept inherited from SDN.MobileFlow has a new entry for supporting the special functions of a mobile network,such as network layer(L3)tun⁃neling and flexible charging.It can also be integrated into OpenFlow networks for basic packet forwarding.A new control⁃ler is abstracted in order to manage new entry⁃based networks, and an OpenFlow controller is introduced to control underlying OpenFlow networks.

In Fig.5,the noticeable enablers are MobileFlow forward⁃ing engine(MFFE)and MobileFlow controller(MFC).MFFEs include all mobile network tunnel processing capabilities. MFFEs also act as wireless access nodes that operate in paral⁃lel with existing eNodeBs in order to manage radio bearers,e. g.,the one near the RAN.MFFEs have more mobile⁃related functions than normal switches,such as OpenFlow switches, and are much simpler than a P⁃GW or router.MFFEs can guarantee QoS and customized services.MFFEs are controlled by an MFC,which is similar to an OpenFlow controller(Fig. 1).MobileFlow also defines a lightweight protocol between the MFC and MFFEs.To support the properties of an existing LTE network,MobileFlow also defines a mobile network application (MNA)north of the control plane.This integrates the functions of existing entities,such as P⁃GW,S⁃GW,and MME.More⁃over,MobileFlow uses OpenFlow networks.In Fig.5,the blue flow goes directly to the Internet through an OpenFlow switch. Such a flow can be offload traffic without QoS requirements. The orange flow passes through a set of MFFEs to a specific service.

Finally,the authors of[8]design a prototype to verify the flexibility of the MobileFlow architecture and show how exist⁃ing mobile networks can evolve through a software⁃defined ar⁃chitecture.

3.2.3 SoftCell vs.MobileFlow

Here,we turn from what SoftCell and MobileFlow support to what they address.Both are based on SDN and provide flexi⁃bility and possibilities for new innovation.Both have separate control plane and data plane.Because they both have a soft⁃ware⁃defined or programmable structure,their architectures can readily be integrated or changed.For example,we can combine MFFEs and OpenFlow switches in MobileFlow:some of these new combined switches will act as access switches and others will act as core switches in SoftCell.An OpenFlow

controller can be merged with an MFC to form a super control⁃ler in SoftCell.

▼Table 1.A service policy

▲Figure 5.MobileFlow architecture.

SoftCell addresses new challenges in software⁃defined archi⁃tecture whereas MobileFlow addresses evolution,i.e.,coexis⁃tence with and gradual replacement of legacy entities.Mobile⁃Flow is designed to benefit operators.The authors of[9]sug⁃gest that operators start to deploy MFFEs,which can operate with legacy equipment via,for example,GTP/PMIP tunnels, and try out new services in the software⁃defined part of the mo⁃bile network.

3.3 Control-Plane Functions in CROWD

CROWD is an SDN project in Europe.At present,the focus of CROWD is the design of control functions in control plane and the mapping of these functions to the north⁃or south⁃end APIs.This ultimately affects physical actions within a network.

CROWD addresses issues in dense heterogonous networks with multiple integrated access networks,such as LTE(macro and small cells)and Wi⁃Fi.In such networks,there are chal⁃lenges in terms of mobility,interference,handover,and energy consumption.To help overcome these challenges,CROWD de⁃fines a series of control functions in an SDN⁃based architec⁃ture.These functions are valuable for future cellular networks. A CROWD controller has a two⁃tier design comprising a CRC for global control and a CLC for local control.The functions of these two controllers are listed in Table 2 and Table 3[10].

4 Discussion

Software⁃defined RAN requires a hierarchal control archi⁃tecture where the local controller can adapt to variable channel conditions.SoftRAN divides the control responsibilities be⁃tween a high⁃level controller and local controller.However,to leverage this principle,researchers also need to analyze more specific cases,i.e.,dynamic traffic(full buffer or non⁃full buf⁃fer).

Moreover,in the RAN information base contained within the high⁃level controller of SoftRAN,weighted interference maps are used to abstract the interference relationship between cells;however,practical interference environments are more complex.To describe the interference space distribution and its dynamics,a multi⁃dimensional interference status space needs to be constructed.Both weighted interference maps and multi⁃dimensional interference status space reflect the physi⁃cal interference conditions through abstraction and are the in⁃put for radio resource management.This modulus can be infor⁃mally regarded as the combination of interference management and radio resource management.Because these functions are implemented in the high⁃level controller with latency consider⁃ations,performance can be improved through prediction or by supporting cognitive interference⁃management schemes based on strategies,Q⁃learning,interference transfer,and avoidance.

The CN still seems to have a long way to go.Even if SoftCelland MobileFlow have some implementation possibilities,opera⁃tors can still not be persuaded to replace their legacy struc⁃tures.After all,these solutions are in the proof⁃of⁃concept stage,and from SoftCell in particular,there are many new chal⁃lenges in software⁃defined cellular networks.Thus,it is better to use an evolving compatible architecture,such as Mobile⁃Flow,at this stage.Meanwhile,researchers need to pay more attention to the uniqueness of and challenges associated with software⁃defined CN as it relates to SoftCell.

▼Table 2.Functions of CRC

▼Table 3.Functions of CLC

One of our research interests is to figure out the effective in⁃terference characteristics approaches,as mentioned previous⁃ly.We are also concerned about the packet⁃scheduling prob⁃lem in software⁃defined cellular networks.By providing fine⁃grained services,the packet may be labeled with the require⁃ments of multiple resource types.For example,one packet needing one CPU time unit and 10 KHz bandwidth to be pro⁃cessed may be labeled with<1,10>as resource profile.Those packets with a multi⁃resource profile will be delivered to par⁃ticularly functional middle⁃boxes(Fig.4).

5 Related Work

In the last two years,there have been many surveys on SDN. In[16]and[17],the authors mainly focus on wireline net⁃

works,present the architecture of SDN,and present the devel⁃opment of each component.OpenFlow is referred to as a spe⁃cial topic in[18].The progress of SDN in wireless networks is referred to in[11]and[19].In[11],the authors divide the tech⁃nique roadmap into wireless WLANs and cellular networks.Al⁃though[11]gives a brief and clear introduction of SDN in wire⁃less networks,it does not elaborate on and compare methods in cellular networks.To the best of our knowledge,this paper is the first work to look into software⁃defined cellular networks. We describe the latest approaches in this field and provide in⁃sights into these approaches.

6 Conclusion and Open Research Areas

In this paper,we have presented state⁃of⁃the⁃art SDN solu⁃tions for cellular networks.Specifically,we have elaborated SoftRAN,which is the latest SDN solution for RAN.SoftRAN has a hierarchical control structure,i.e.,the high⁃level control⁃ler makes decisions based on global information,and the local controller can adapt to variable channel conditions.SoftRAN is based on the principle of dividing control responsibilities be⁃tween a high⁃level controller and local controllers.SoftCell and MobileFlow are CN⁃related SDN solutions.We have compared these approaches and provided insight into the design of soft⁃ware⁃defined cellular networks.We have also discussed our re⁃cent work in this direction and expect to invoke some new ideas on software⁃defined cellular networks.

There are still many research areas that can be exploited. Mobile networks tend to be dense and large⁃scale.Although SoftCell and CROWD address some issues,their performance in particular cases is still unclear.To increase capacity,a dis⁃tributed massive MIMO system is necessary in the future.SDN⁃enabled cross⁃layer MIMO is attractive,especially for the con⁃figuration of different beamforming matrixes.Another interest⁃ing area is combined use of different access techniques.Exist⁃ing solutions are still in the proof⁃of⁃concept stage.

[1]N.Feamster,J.Rexford,and E.Zegura.The Road to SDN:an intellectual history of programmable networks[Online].Available:https://www.cs.princeton.edu/ courses/archive/fall13/cos597E/papers/sdnhistory.pdf

[2]N.McKeown,T.Anderson,H.Balakrishnan,G.Parulkar,L.Peterson,J.Rex⁃ford,S.Shenker,and J.Turner,“Openflow:enabling innovation in campus net⁃works,”ACM SIGCOMM Computer Communication Review,vol.38,no.2,pp. 69-74,2008.doi:10.1145/1355734.1355746.

[3]K.K.Yap,M.Kobayashi,R.Sherwood,T.Y.Huang,M.Chan,N.Handigol,and N.McKeown,“Openroads:Empowering research in mobile networks,”ACM SIG⁃COMM Computer Communication Review,vol.40,no.1,pp.125-126,2010. doi:10.1145/1672308.1672331.

[4]K.K.Yap,R.Sherwood,M.Kobayashi,T.Y.Huang,M.Chan,N.Handigol,N. McKeown,and G.Parulkar,“Blueprint for introducing innovation into wireless mobile networks,”in Proceedings of the second ACM SIGCOMM workshop on Vir⁃tualized infrastructure systems and architectures,New York,USA,pp.25-32. doi:10.1145/1851399.1851404.

[5]L.Suresh,J.Schulz⁃Zander,R.Merz,A.Feldmann,and T.Vazao,“Towards pro⁃grammable enterprise WLANs with Odin,”in Proceedings of the first workshop on Hot topics in software defined networks,New York,USA,pp.115-120.doi: 10.1145/2342441.2342465.

[6]L.E.Li,Z.M.Mao,and J.Rexford,“Toward software⁃defined cellular net⁃works,”in Software Defined Networking(EWSDN),2012 European Workshop on, pp.7-12.doi:10.1109/EWSDN.2012.28.

[7]A.Gudipati,D.Perry,L.E.Li,and S.Katti,“SoftRAN:Software defined radio access network,"in ACM SIGCOMM HotSDN Workshop,New York,USA,pp. 25-30.doi:10.1145/2491185.2491207.

[8]X Jin,L.E.Li,L.Vanbever,and J.Rexford.SoftCell:Taking control of cellular core networks[Online].Available:http://arxiv.org/abs/1305.3568

[9]K.Pentikousis,Yan Wang,and Weihua Hu,“Mobileflow:Toward software⁃de⁃fined mobile networks,”Communications Magazine,IEEE,vol.51,no.7,pp. 44-53,2013.doi:10.1109/MCOM.2013.6553677.

[10]H.Ali⁃Ahmad,C.Cicconetti,A.de la Oliva,et al.,“CROWD:An SDN Ap⁃proach for DenseNets,”Software Defined Networks(EWSDN),2013 Second European Workshop on,vol.12,no.2,pp.25-31,2013.doi:10.1109/EWS⁃DN.2013.11.

[11]Qadir Junaid,Nadeem Ahmed,and Nauman Ahad.Building Programmable Wireless Networks:An Architectural Survey[Online].Available:http://arxiv.org/ abs/1310.0251

[12]N.Gude,et al.,“NOX:Towards an Operating System for Networks”,SIG⁃COMM CCReview,Vol.38,Issue 3,July 2008.doi:10.1145/1384609.1384625.

[13]Floodlight Openflow Controller[Online].Available:http://floodlight.openflow⁃hub.org/

[14]E.Dahlman,S.Parkvall,and J.Skold,4G LTE/LTE⁃Advanced for Mobile Broad⁃band,UK:Academic Press,2011,pp.95-127.

[15]Cisco PGW Packet Data Network Gateway[Online].http://www:cisco:com/en/ US/products/ps11079/index:html

[16]M.Mendonc,B.N.Astuto,X.N.Nguyen,K.Obraczka,T.Turletti,et al..A sur⁃vey of software⁃defined networking:Past,present,and future of programmable networks[Online].Avalibale:http://hal.inria.fr/hal⁃00825087/

[17]K.Hyojoon and N.Feamster,“Improving network management with software de⁃fined networking,”Communications Magazine,IEEE,vol.51,no.2,pp.114-119,2013.doi:10.1109/MCOM.2013.6461195.

[18]A.Lara,A.Kolasani,and B.Ramamurthy,"Network Innovation using Open⁃Flow:A Survey,"Communications Surveys&Tutorials,IEEE,vol.16,no.1, pp.493-512,2013.doi:10.1109/SURV.2013.081313.00105.

[19]S.Costanzo,L.Galluccio,G.Morabito and S.Palazzo,“Software Defined Wire⁃less Networks:Unbridling SDNs,”in Software Defined Networking(EWSDN), 2012 European Workshop on,Darmstadt.doi:10.1109/EWSDN.2012.12.

Manuscript received:2014⁃02⁃15

BiograpphhiieessJiandong Li(jdli@mail.xidian.edu.cn)received his BE,MS,and PhD degrees in Communications Engineering from Xidian University,Xi’an,in 1982,1985 and 1991.He has been a faculty member of the school of Telecommunications Engineer⁃ing,Xidian University,since 1985.He is currently a professor and vice director of the academic committee of the State Key Laboratory of Integrated Service Networks. Professor Li is a senior member of IEEE.He was a visiting professor in the Depart⁃ment of Electrical and Computer Engineering,Cornell University,from 2002 to 2003.He was the general vice chair of ChinaCom 2009 and TPC chair of IEEE ICCC 2013.He was awarded as Distinguished Young Researcher award from NSFC and Changjiang Scholar from the Ministry of Education,China.His main research interests include wireless communication theory,cognitive radio,and signal process⁃ing.

Peng Liu(liupeng0218@gmail.com)is a PhD candidate at Xidian University, Xi’an,China.He is also a visiting scholar at Columbia University,NY.He received his BS degree in Telecommunications Engineering from Xidian University in 2010. His research interests include resource allocation,interference avoidance,and inter⁃ference cancellation in heterogeneous networks;and packet scheduling in SDN.

Hongyan Li(hyli@xidian.edu.cn)received her MS degree in control engineering from Xi’an Jiaotong University,China,in 1991.She received her PhD degree in sig⁃nal and information processing from Xidian University,China,in 2000.She is cur⁃rently a professor in the State Key Laboratory of Integrated Service Networks,Xidi⁃an University.Her research interests include wireless networking,cognitive net⁃works,integration of heterogeneous network,and mobile ad hoc networks.