A 5G Virtualized RAN Based on NO Stack

Jie Zeng, Xin Su, Jinjin Gong, Liping Rong, Jing Wang

Tsinghua National Laboratory for Information Science and Technology, Research Institute of Information Technology,Tsinghua University, Beijing 100084, China

* The corresponding author, email: suxin@tsinghua.edu.cn

I. INTRODUCTION

The 5G mobile communication system has brought numerous particular requirements and challenges [1], such as the ultra-dense cellular network [2], the millimeter wave communications based network densification [3], and how to balance the vehicular communication capacity and the handoff ratio in 5G [4]. Thus,5G is calling for the novel network architecture design and implements. One of the promising network technologies is network virtualization, which abstracts the common underlying infrastructure and provides a unified and programmable interface through virtualization technologies. When the network virtualization is firstly applied to the wired network, this technology can build networks quickly and easily, reduce the network construction investment, and simplify the network maintenance.Recently, it has been employed in the wireless network and could achieve the similar benefits.

Several network function virtualization(NFV) and software-defined network (SDN)related solutions have been proposed in recent research. A novel network architecture is presented in [5], consisting of application cloud,SDN controller cloud, SDN based Cloud-RAN(C-RAN), SDN based transport network, and SDN based core network. In [6], the distributed resource collecting process in the SDN is studied. In [7], a wireless software-defined cellular system (SoftAir) is firstly proposed with details of the system architecture, management tools, and traffic engineering solutions. Wireless software-defined networks (W-SDNs) for 5G are proposed in [8], and the fundamentals,state-of-the-art and limitations of the W-SDN have been studied. A SDN based carrier grade Wi-finetwork framework (SWN) is proposed in [9] to divide Wi-finetworks into control and data planes, thereby supporting seamless handover between APs, automatic network access and unified authentication. Virtual Network Operators (VNOs), whose services are provided by virtual resources, use wireless connectivity to make the RAN more flexible and efficient [10]. Virtual RATs adopt protocol stacks with software-based and programmable features, which could run on the same hardware platform [11]. Network virtualization substrate (NVS), which is based on spectrum sharing, is designed as a hierarchical scheduler to provide customization and isolation [12].Taking advantages of general-purpose platforms, China Mobile Communications Corporation (CMCC) built a pre-commercial Mini C-RAN systems to deal with the challenge of high-density RRU flexible networking in Mobile World Congress 2016. The key technologies involved in the evolution of C-RAN are discussed in [13], Frameless Network Architecture (FNA) is proposed to form a cell-lessbased network.

However, the solutions mentioned above have inherent shortages and are not efficient enough. Meanwhile, the vRAN based technology could improve the network performance significantly. The vRAN, following the trend of the virtualization in wireless networks, is widely accepted among mobile operators. The main challenge is trying to virtualize baseband physical layer (PHY), which contains the most computational intensive signal processing tasks. A use case of “Virtual Base Station(VBS) L1 Acceleration” is described in [14].The computing, storage and networking resources are abstracted from specific hardware resources and implemented on a general-purpose platform.

In this article, a vRAN based on NO stack is analyzed and demonstrated. NO Stack adopts virtualization principles into cellular networks, pushes down the vertically stacked protocol stacks and encapsulates protocol layers as modules. The protocol modules, as the virtual resources, can be managed and orchestrated to implement an integral RAN with the corresponding protocols [15],[16]. A setup approach of the NO Stack based vRAN is described in our prior work [17] in detail. The remainder of this paper is organized as follows.Section II describes the overall framework of NO Stack and introduces the components of the vRAN based on NO Stack. Section III presents the key concepts of the RAN slice and network orchestration schemes. In Section IV, the use cases of dedicated bearer and default bearer establishments based on NO Stack are introduced and analyzed. In Section V, demonstrations and measurements show the benefits brought by the proposed vRAN approach. Finally, conclusions and future works are given in Section VI.

II. THE VRAN BASED ON NO STACK

As shown in Fig. 1, the vRAN based on NO Stack is divided into the management plane(M-plane), control plane (C-plane) and user plane (U-plane). Therefore, it has the ability to provide features like the intelligent management, flexibility and abstraction. The M-plane is formed of the infrastructure and network logic resources, such as the coverage scheduler, policy setting, and network orchestration, to implement customized services. The C-plane provides logically centralized control functions and supports the M-plane via an open interface between distributed network elements. With access nodes, cloud nodes,networking nodes and associated links inside,the U-plane could provide functions (such as data transmission and processing) in a distributed and flat style. The processing procedures in the U-plane are configurable and programmable, while the delay of interactions with the C-plane can be reduced markedly. The control functions and orchestration are based on the global network view (GNV) which is placed in the C-plane. GNV takes control of multiple wireless access node resources and centralizes them in the virtual resource pool.

The virtual resource pool of the vRAN replaces the traditional telecommunication dedicated hardware with general-purpose platforms formed by physical and software resources. It contains a general processing platform, virtualization layer, multiple virtual network functions (VNFs), management and orchestration (M&O), and RAN slices, which could be realized by organically aggregating the other mentioned components. The virtualization layer is able to abstract and partition physical resources logically. Then it could decouple VNF software from the underlying hardware. While enabling VNFs to use the underlying virtualized infrastructure, the virtualization layer could supply numerous virtual resources. The protocol module pool consists of loosely coupled virtual protocol modules.In most scenarios, the virtual protocol modules could be orchestrated and configured intelligently in time. And they could also keep updating to the latest version in an efficient way. Since a VNF is a concrete software realization of different network functions, a RAN slice could be divided into one or more VNFs which should be connected through a unified interface.

Fig. 1 The block diagram of the vRAN based on NO stack

III. KEY TECHNOLOGIES OF THE VRAN

3.1 The RAN slice

The RAN slice consists of a set of RAN functions, which have a large number of logical and physical resources. A RAN slice provides users with the agile and efficient access service [18]. Various RAN slices can be created by plentiful styles to support different features of various existing and upcoming RATs. The concept of the RAN slice is illustrated in Fig.2.

As mentioned above, there are three logical planes in the vRAN based on NO Stack. The relevancies between these planes and the RAN slice are described as follows.

In different radio access scenarios, the M-plane supports the RAN slice by dynamically creating RAN slices, which could be organized by different independent functions and be configured by the corresponding parameters. When an operator or a third party offers the access service, it should be specified by an access service instance. To meet the requirements of an access service instance, a RAN slice blueprint enables the illustration of the specific RAN slice.

The slice selection is placed in the C-plane and helps UEs access dedicated RANs. UEs should be steered to different RAN slices according to selected access technologies.

There are numerous function modules located in the U-plane. These function modules can be dynamically assembled to complete the forwarding function of a RAN slice, so the forwarding function could support user data processing in different RATs. At the same time, user flows can be steered to the corresponding function modules to accomplish a specific procedure according to the selected RAN slice.

3.2 The network orchestration

A network orchestrator plays a major role in the network orchestration. It is the key to translating the use cases and service models into actual network functions. A network orchestrator is responsible for the lifecycle management of the VNF instances, as well as RAN slices and protocol modules.

In the vRAN based on NO Stack, all protocols are packaged as modules, in which flexible and programmable implementations could be achieved. With the help of GNV, the network orchestrator can dynamically orchestrate the proper protocol modules to formula a certain function. The loose coupling among protocol modules allows the orchestrator to avoid some unnecessary modules, thereby reducing the time of data processing.

The network orchestrator operates following the instructions from the upper layer. It firstly decides the RAN slice for a specific use case. Then the network orchestrator connects the relevant VNFs and protocol modules.After the network orchestrator distributes the relevant configurations based on the QoS requirement, it finally maps the above mentioned functions to the infrastructure resources. During the RAN slice lifecycle, the orchestrator should monitor and adjust RAN slices up to time to meet the various resource requirements. Moreover, the priority and fairness among different RAN slices should be considered in the resource allocation.

IV. USE CASES AND ANALYSES

NO Stack is still in the development stage,and some vRAN use cases based on NO Stack are studied in this section. Observed from the dedicated and default bearer establishment procedures based on NO Stack, the signaling overheads and service response time can be markedly reduced.

4.1 The dedicated bearer establishment

NO Stack considerably redefines the traditional Network Elements. In this innovative framework, the global controller consists of the functions of mobility management entity(MME), mobile anchoring, domain name system (DNS), policy and charging rule function(PCRF), while setting the flow table. A flow table control gateway (GW) is contained in xGW. However, the tunnel is no longer needed and the GW function is removed with process-ing capability reserved for the eNodeB. Then,we will compare dedicated bearer establishment procedures between LTE and NO Stack.

Fig. 2 An illustration of the RAN slice

A typical end-to-end dedicated bearer establishment procedure is depicted in Fig. 3.The setup procedure for LTE bearer service includes the orderly established radio bearer,S1 bearer, and S5/8 bearer [19]. This sequence segment mechanism of bearer establishment causes a long delay in the construction of the link, and also risks other relevant segments by reflecting the exception, resulting in failures.

Fig. 3 The dedicated bearer establishment procedure in LTE/SAE

Fig. 4 The dedicated bearer establishment procedure in NO Stack

In NO Stack, this dedicated bearer establishment procedure can be reconstructed as follows. 1) Bearer establishment needs one request for attaching; 2) All the states of the relevant sessions in global conversation view can be modified at the same time; 3) All the relevant processing modules in the U-plane can be programmed at the same time; 4) The whole procedure above is packaged as an atom transaction and it will be totally recreated if abnormal.

The dedicated bearer establishment procedure of NO Stack is shown in Fig. 4 and it only includes two bearer procedures. In NO Stack, the steps 1-4 in Fig. 3 are no longer executed continuously but simultaneously in the eNodeB, GNV, xGW in parallel under the help of the global controller. Certainly, the steps 7-10, which are the corresponding response messages to confirm that the bearers have been correctly set up, could be no longer transmitted independently.

Thus, the delay of building links and signaling overheads can be significantly reduced.In an assumed successful process of the bearer establishment, there are no more than 8 pieces of signaling in NO Stack while at least 10 pieces of signaling in LTE, which means a 20 percent reduction of signaling overheads and more signaling overheads can be reduced in extraordinary cases. Moreover, as for every protocol layer or every control procedure, NO Stack provides transaction services instead of a separated mechanism for exceptions. It will greatly simplify the design and protocol implementation.

4.2 The default bearer establishment

The default bearer firstly connects to a public data network (PDN), which is in charge of assigning an IP address to the UE and ensuring that UE is always connected with the assigned IP. The default bearer also remains established throughout the lifetime of the connection. However, the dedicated bearer can be established during or after the completion of the connection to PDN. Therefore, the major difference between the default bearer and dedicated bearer in the access network is transparent.

The delay budget of the default bearer establishment between different entities in LTE is given in Table 1. According to reconstruction design rules of NO Stack and the bearer establishment, the default bearer establishment based on NO Stack also undertakes a corresponding change similar to the dedicated bearer establishment.

The procedure of MME to DNS and that of packet gateway (PGW) to PCRF will change to process inside a unique network global controller. Meanwhile, service gateway (SGW)and PGW only remain respective flow table control functions and unify them as xGW does. Therefore, the corresponding delay from SGW to PGW can be ignored. Finally, it is observed that in the default bearer based on NO Stack, the establishment delay can be reduced by at least 300 ms.

As shown in Table I, the total time for the bearer establishment is about 500 ms in the existing LTE system. With nonstop development of technologies, the processing capacity and the processing delay of each entity also continue to improve obviously. As a consequence,the default bearer setup delay based on NO Stack can be reduced by at least 60 percent.

V. DEMONSTRATIONS AND MEASUREMENTS

A demonstration is built to verify and evaluate the vRAN based on NO Stack. Totally 3 outdoor macro cells and 18 indoor micro cellswith 60 antennas are deployed on different floors. The infrastructures are integrated by the OpenStack technology, and customized according to requirements of actual applications. The management platform is deployed in a dedicate room, with the management instructions being transmitted to the eNodeB,MME, SGW and PGW follow the pattern of NO Stack. This platform of concepts could implement the virtualization of the infrastructure resources to enable the highly flexible of the network and seamless migration.

Table I Delay budget for the default bearer establishment

A test based on the heterogeneous RATs fusion is designed and carried out for the proposed vRAN. The core technical indicators are measured and evaluated, including signaling overheads and service response time between heterogeneous RATs. Required equipment and test resources are adopted in the assessment process, such as, multi-mode terminals including mobile phones, test computers matched the terminals, and a signaling analyzer used to track the signaling process.

A call between heterogeneous RATs in existing LTE network is shown in Fig. 5 and the one in the proposed NO Stack based vRAN is shown in Fig. 6. The specific process could be described in details as follows. At first, a LTE user initiates a call and Multi Media Telephony (MMTel) Application Server (AS) performs caller service processing. Then MMTel AS routes the call back to Serving-Call Session Control Function (S-CSCF), which queries Telephone Number Mapping (ENUM) and routes the call to the called 2G/3G user. On the other side, the call is routed to the Interrogating-Call Session Control Function (I-CSCF),which sends request to the S-CSCF. After MMTel AS calls the call back to S-CSCF,S-CSCF triggers Service Centralization and Continuity (SCC) AS to select a LTE access network the called 2G/3G user. When S-CSCF connects the call request to the called user through Proxy-Call Session Control Function(P-CSCF), the bearer establishment achieves.

Fig. 5 The test diagram of heterogeneous RATs in existing LTE

Fig. 6 The test diagram of heterogeneous RATs in the NO Stack based vRAN

Evaluation results show that the proposed vRAN outperforms the existing LTE solution in the scenario of heterogeneous RATs. As expected, it reduces average signaling overheads by about 21.9 percent. Service response time between heterogeneous RATs considerably changes in different transmitting environments. It is around 80 percent lower than that of the CSFB in the idle state, and is more than 60 percent lower in the data connection state.

VI. CONCLUSIONS AND FUTURE WORK

It is difficult for the existing RAN architecture to completely satisfy the requirements of the 5G. The vRAN has been well known as the trend and aroused many research interests.In this paper, the concept of NO Stack has been approved in the vRAN to make it more flexible, programmable and sustainable. The RAN slice and network orchestration schemes have been deployed in the proposed vRAN to form different RAN slices, thereby supporting multi-RATs and customized services. Meanwhile, the proposed vRAN could virtualize network components and change specific procedures such as the bearer establishment.Some use cases and analyses have been discussed to verify the advantages of the NO Stack based vRAN. Finally, a test based on the heterogeneous RATs fusion has been designed and carried out as a demonstration of the proposed vRAN, and some meaningful measurement results have been obtained. As a result,signaling overheads in the dedicated bearer establishment could be reduced by more than 20 percent. Besides, service response time between heterogeneous RATs might be reduced by at least 60 percent as well.

There are also some fundamental problems needed to be addressed in the future, such as the parallel processing, global data updating and function split granularity. More simulations and demonstrations will be realized to help optimize the vRAN based on NO Stack.With the continuing development, the vRAN scheme could be a promising technology of the 5G network architecture. And the proposed vRAN based on NO Stack enables the scalable capacity of the next generation network to satisfy the various KPIs of the characteristic scenarios, especially the critical machine type communications.

ACKNOWLEDGEMENTS

This work was supported by the China's 863 Project (No. 2015AA01A706), the National Science and Technology Major Project (No. 2016ZX03001017), the Science and Technology Program of Beijing(No. D161100001016002), and the Science and Technology Cooperation Projects (No.2015DFT10160B).

[1] S. Chen and J. Zhao, “The requirements, challenges, and technologies for 5G of terrestrial mobile telecommunication”,IEEE Commun.Mag.,vol. 52, no. 5, pp. 36–43, 2014.

[2] X. Ge, S. Tu, G. Mao, C. X. Wang, and T. Han,“5G Ultra-Dense Cellular Networks”,IEEE Wirel.Commun.,vol. 23, no. 1, pp. 72–79, 2016.

[3] W. Feng, Y. Wang, D. Lin, N. Ge, J. Lu, and S. Li,“When mmWave communications meet network densification: a scalable interference coordination perspective”,IEEE Journal on Selected Areas in Communications, Apr. 2017.

[4] X. Ge, H. Cheng, G. Mao, Y. Yang, and S. Tu,“Vehicular Communications for 5G Cooperative Small-Cell Networks”,IEEE Trans. Veh. Technol.,vol. 65, no. 10, pp. 7882–7894, 2016.

[5] Z. Ma, Z. Zhang, Z. Ding, P. Fan, and H. Li, “Key techniques for 5G wireless communications:network architecture, physical layer, and MAC layer perspectives”,Sci. China Inf. Sci., vol. 58,no. April, pp. 1–20, 2015.

[6] X. Wu, C. Wu, C. Lin, Q. Wu, and B. Wang, “A multipath resource updating approach for distributed controllers in software-defined network”,Sci. China Inf. Sci., vol. 59, no. 9, pp. 0–10,2016.

[7] I. F. Akyildiz, P. Wang, and S. C. Lin, “SoftAir: A software defined networking architecture for 5G wireless systems”,Comput. Networks, vol. 85,pp. 1–18, 2015.

[8] I. F. Akyildiz, S. C. Lin, and P. Wang, “Wireless software-defined networks (W-SDNs) and network function virtualization (NFV) for 5G cellular systems: An overview and qualitative evaluation”,Comput. Networks, vol. 93, pp. 66–79,2015.

[9] L. Tao, X. Wen, Z. Lu, X. Zhao, Y. Li, and Zhang,Biao, “SWN: An SDN Based Framework for Carrier Grade Wi-fiNetworks”, [J]China Commun., vol. 13, no. 3, pp. 12–26, 2016.

[10] S. Khatibi, L. M. Correia, “Modelling of virtual radio resource management for cellular heterogeneous access networks”, IEEE,International Symposium on Personal, Indoor, and Mobile Radio Communication,pp.1152-1156 2014.

[11] S. Chen, J. Zhao, M. Ai, D. Liu, and Y. Peng,“Virtual RATs and a flexible and tailored radio access network evolving to 5G”,IEEE Commun.Mag.,vol. 53, no. 6, pp. 52-58, 2015.

[12] X. Costa-Perez, J. Swetina, T. Guo, R. Mahindra,and S. Rangarajan, “Radio access network virtualization for future mobile carrier networks”,IEEE Commun. Mag.,vol. 51, no. 7, pp. 27-35,2013.

[13] X. Xu, Z. Sun, X. Dai, X. Tao, and P. Zhang, “A Frameless Network Architecture for the Way Forward of C-RAN”, [J]China Commun., vol. 13,no. 6, pp. 154–166, 2016.

[14] ETSI, “Network Functions Virtualisation (NFV);Acceleration Technologies; Report on Acceleration Technologies & Use Case”, V1.1.1, Dec.2015.

[15] X. Mi, Z. Tian, X. Xu, M. Zhao, and J. Wang, “NO stack: A SDN-based framework for future cellular networks”,IEEE Wireless Personal Multimedia Communications (WPMC), pp.497-502, Sept.2014.

[16] J. Zeng, L. Rong, and X. Su, “NO stack: A software defined framework for 5G mobile network”,IEEE International Conference on Telecommunications, pp.1-5, May 2016.

[17] J. Zeng, X. Su, J. Gong, L. Rong and J. Wang,“5G virtualized radio access network approach based on NO Stack framework”,IEEE Int. Conf.Commun., pp.1-5, May 2017.

[18] 3GPP, “Study on Architecture for Next Generation System (Release 14)”, Technical Specification Group Services and System Aspects, vol. TR 23.799 V0.7.0, Aug. 2016.

[19] 3GPP, “General Packet Radio Service (GPRS)enhancements for Evolved Universal Terrestrial Radio Access Network (E-UTRAN) access”, Technical Specification Group Services and System Aspects, vol. TS 23.401 V14.1.0, Sept. 2016.