Abstract
Focusing on the layout of the 5G mobile communication base station in the city center, we design a 5G city network slicing strategy for the three typical application scenarios with enhanced mobile broadband (eMBB), ultrareliable low-latency communications (URLLC), and massive machine type communications (mMTC). The strategy considers multiple important network performance indicators, including user guaranteed bandwidth, maximum bandwidth limit, QoS (quality of service), link delay tolerance, and slicing throughput. The slicing strategy can greatly increase the connections of base station clients and the utilization of network resources, and effectively reduce block radio and handover radio. The simulation experiments adopt the 5G base station dataset of a coastal city layout in Zhejiang province. Our tests show that the 5G network slicing strategy has certain advantages in network transmission performance in urban complex environment. The research can provide an effective reference for 5G infrastructure construction in other cities.
1. Introduction
The fifth generation mobile communication technology (5G) was launched by the European Union in 2013 with huge investment, and now it has been put into the application and officially entered the market [1]. The design, goal, and the most significant feature of 5G is to meet users’ requirements for speed, time delay, reliability, traffic density, and so on. The corresponding application scenarios can be subdivided into enhanced mobile broadband (eMBB), ultrareliable low-latency communications (URLLC), and massive machine type communications (mMTC). Research focused on the above three application scenarios [2] can greatly improve the mobile communication experience of individual users, promote the derivative development of more emerging industries [3], and finally promote innovative fields such as digital city [4], automatic driving, smart medical care, and bring about industrial revolution [5]. Therefore, 5G can be regarded as a forward-looking information infrastructure construction [6], providing basic technical support [7], and real experimental environment for the future internet of everything [8].
According to the data provided by the Ministry of Industry and Information Technology, China is expected to build more than 600,000 5G base stations by the end of 2020, covering all cities at prefecture level and above. Therefore, in the case of the comprehensive development of 5G network construction in domestic cities [9], priority should be given to the environment of urban centers to optimize the 5G ecological network, realize the comprehensive deployment of 5G backbone network, and help the safe landing of the three application scenarios of eMBB, URLLC, and mMTC [10]. The eMBB is generally used for high-volume mobile broadband services, such as 3D ultrahigh-definition video. The URLLC is generally used for low-latency services, such as autonomous driving and industrial automation. In addition, the mMTC is commonly used for IoT device deployment, such as wireless sensor networks. To meet 5G city under complex environment the user demand for customized logic network and network virtualization [11], and ensure the high performance of transmission delay, peak rate, and security, network slicing is considered a key technology to solve these problems, has also been recognized as the ideal of 5G network architecture by both industry and academia [12]. The essence of network slicing is a proprietary network that shares the underlying common infrastructure but is logically completely isolated in the upper layer [13], which can be dynamically combined according to business requirements and finally generate edge network slices [14]. To sum-up, slice-oriented research on 5G urban complex network strategy is at present of great significance for the construction of 5G base stations and the corresponding applications [15].
2. Related Works
Network slicing can provide isolated network environment for different application scenarios, and each application scenario can be customized according to its own business requirements. 5G physical network can be divided into multiple isolated logical network [16] by using network slicing, in order to realize efficient utilization of network resources such as device end, access network, backbone network, system operation, and maintenance. Because network slicing technology has absolute advantages in the field of 5G communication, researchers around the world have carried out related studies on 5G network slicing in recent years. He et al. [17] proposed a slicing virtual network functions (VNF) strategy for 5G Network based on security threat prediction by analyzing the lag of existing VNF in response to coresident attacks. Jiang et al. [18] proposed an admission control mechanism based on heuristic algorithms, which can realize dynamic allocation of network resources by slicing and meet users’ demand for network bandwidth to the greatest extent. Tang et al. [19] aimed at the VNF migration optimization problem under 5G network slicing architecture, based on the stochastic optimization model of constrained Markov decision process, realized the dynamic deployment of multiple service function chain, and reduced the average energy consumption of infrastructure. Isotalo considers that there must be a large number of malicious network attacks on the chain after the popularity of 5G network, so it designs a network security model to isolate the user equipment (UE) to a dedicated test network slice to realize the detection and analysis of UE traffic [20].
According to the above, at present, domestic and foreign scholars mainly focus on the research in the field of 5G network slicing security, but seldom involve the application of network environment in actual deployment. The most critical point is that no researchers have yet analyzed the reasons for network congestion, and the impact of network traffic characteristics for 5G network transmission quality. Therefore, the study combined with the existing 5G network slicing technology, fully considering the characteristics of high-base station construction density and large person traffic in the urban center environment, and designed a 5G network slicing strategy for the complex urban environment. This strategy can ensure the maximum user bandwidth, link delay, data throughput, and other network performance, at the same time, improve the number of base station user connections and network resource utilization, reduce frequent blocking and switching times of users, and provide reference for the further implementation of 5G network in the future.
3. 5G Network Transmission Process and Slicing Strategy
3.1. Analysis of 5G Network Transmission Process
When the initial access of 5G NG Radio Access Network (NG-RAN) is on, the UE first needs to find 5G wireless network and establish network connection, including the following two steps. Step 1. Obtaining synchronization of uplink (UL) and downlink (DL), namely, obtaining DL synchronization through listening network, and acquiring UL synchronization through random access.
Step 2. Sending and receiving messages, establishing connections. It includes setting up mobility context, default load, and attachment process between UE and 5G core network to obtain IP address assigned by network.
In the nonstandalone (NSA) networking mode during the initial access process, the NG-RAN base station, or gNodeB, does not need to broadcast remaining minimum system information (RMSI). The contents in RMSI are sent to the UE by radio resource control (RRC) signaling sent by LTE before the UE begins to connect to the new radio (NR). Among them, gNodeB can be separated as the centralized unit (CU) and the distributed unit (DU) architecturally, so there are two types of base stations: gNB-DU and gNB-CU. The 5G wireless access process is shown in Figure 1.
NG-RAN includes two types of nodes: the first type is gNB node [21], which can provide NR user plane and control plane protocol terminals to UE; the second type is ng-eNB node [22], which can provide user plane and control plane protocol terminals in evolutionary land wireless access network to UE. The gNB and ng-eNB can communicate each other through the network interface between NG-RAN nodes. gNB and ng-eNB are also connected to the 5G core network through the NG interface, namely, they can be connected to the access and mobility management function (AMF) and user plane function (UPF) through the interface. The session management function (SMF) is responsible for assigning and managing IP addresses of UE, selecting and controlling UP functions, configuring the traffic routing of UPF, controlling instruction execution and QoS policies, and downlink data notification. The overall architecture of RAN is shown in Figure 2.
3.2. Design of 5G Network Slicing Strategy
The core of 5G network slicing strategy design is to provide customized network characteristics, such as bandwidth, latency, and capacity, while satisfying multiconnection, multiservice, and flexible deployment. Multiple logical networks can be segmenting from an independent physical network through network slicing technology. The segmentation of wireless resources of 5G base station is shown in Figure 3.
The 5G communication standards have been continuously updated and iterated. The latest progress of 5G technology is that in 2022, the 5G R17 standard has been announced to be frozen by the international communication standard organization 3GPP. The 5G R18 standard is expected to be frozen by the end of 2023. The 5G network slicing strategy we designed, including normalization and information entropy calculation, is a specific plan based on the R17 standard and specific technical means.
Through in-depth investigation and research [23], the following five network performance indicators are selected as the reference basis for 5G network slicing strategy in complex urban environment, as shown in Figure 4. The parameters used in the existing network slicing strategy are used in the construction of network slicing to calculate the weight. The main purpose is to satisfy the maximum QoE (quality of experience) of all users. Therefore, the quality of experience of each user will be greater than the overall average quality of service (QoS) in practice. However, we consider optimizing the overall QoS performance of the entire network, and therefore chose five representative network performance parameters, including bandwidth usage, maximum bandwidth limit, throughput, QoS class, and link delay tolerance.
3.2.1. Bandwidth Usage
As 5G adopts more advanced symbol forming technology, it reduces the overhead of spectral edge protection band. Therefore, compared with the traditional 4G network, the transmission bandwidth has been greatly improved. However, in mMTC application scenarios, 5G network usually require higher bandwidth for per unit area. Considering the huge person traffic in a complex urban environment, there will be a large number of devices in a specific area that request connection at the same time.
3.2.2. Maximum Bandwidth Limit
In the actual deployment of eMBB application scenario, the radio resource of 5G base station is usually constant without considering base station reconstruction or subsequent optimization. At this time, it is necessary to set the maximum bandwidth limit to ensure that one user will not preempt too many resources, which will lead to the bandwidth limitation of other users and reduce the overall user experience.
3.2.3. Throughput Occupancy
For interactive application network, throughput index reflects the pressure that the whole network can bear. As a more intuitive network performance index, throughput can reflect the impact of slicing on system load capacity. Throughput selection slice can provide the maximum data transmission rate for users in slice access service, and realize the handover process control under 5G network slicing.
3.2.4. QoS Class
According to the definition of 3GPP standard specification, QoS is divided into five class types: (1) real-time conversation class, such as voice over Internet protocol (VoIP), which is widely used; (2) streaming class, such as video; (3) interactive class, including the popular Internet, such as web browsing; (4) background class, such as email, file transfer protocol (FTP), and short message service (SMS); (5) in addition, other types can be treated as best effort flow class. QoS class is shown in Table 1.
3.2.5. The Link Delay Tolerance
Since the beginning of 5G design, the network delay characteristics have become a part of 5G requirements. The ultralow delay and ultrahigh reliability application scenarios in URLLC lead to higher requirements for 5G network round-trip delay, which usually needs to reach less than 1 MS, which is nearly 10 times lower than the 10 MS required by 4G network. Therefore, when slicing the network, the tolerable delay of the link should be taken into account.
3.3. 5G Network Slicing Weight Calculation
Based on the above analysis of 5G network slicing strategy, the impact of five indicators on 5G network slicing performance in complex urban environment can be mainly considered: bandwidth occupancy, maximum bandwidth limit, single slice occupancy throughput, QoS class, and link tolerable delay. Therefore, in order to better model and analyze the network slicing strategy, the above performance indicators can be quantified by five network data, namely guaranteed bandwidth (single user), maximum bandwidth (single user), guaranteed throughput (single slice), QoS priority, and maximum delay (single user). Entropy weight method can be used to calculate the slicing weight matrix under the influence of five network performance indexes, and then the slicing weight vector can be determined finally.
In order to compare the performance indicators of network slicing under the same metric standard, it is necessary to normalize the starting matrix of slicing. By normalizing the vector of slicing performance index, the slicing normalization matrix can be obtained, which can be used to calculate the weight matrix. The initial matrix X and the normalized matrix R are shown in Equations (1) and (2), respectively, and the specific calculation process of vector normalization is shown in Equation (3).
The parameter xij is the original value of the j-th performance index in the i-th slice of the starting matrix X, and rij is the normalized value of the j-th performance index in the i-th slice of the normalization matrix R. The parameter of m and n are the total number of slice and performance indexes.
In order to consider the relative importance of slice performance index, the weighted normalization matrix V obtained can be calculated by weighting the normalization matrix, as shown in Equation (4).where wj is the weight of the j-th performance index, and vij is the weighted normalized value of the corresponding j-th performance index in the i-th slice of the weighted normalization matrix V.
Entropy weight method adopted is based on probability theory as the basis of data uncertainty measurement, which indicates that wide data distribution will bring higher uncertainty [24]. The information entropy of each performance index in the slice initiation matrix can be used for accurate weight calculation according to the relative difference between slices. The calculation process of information entropy ej of slicing performance index is shown in Equations (5) and (6).
The weight of slicing performance index wj can be calculated through information entropy ej, as shown in Equation (7) as:
After the performance index weight wj is calculated according to Equation (7), the weighted normalization matrix V is brought into Equation (4), which can be used as the slicing weight matrix. The weighted normalization values of each performance index are accumulated through calculation, and the final proportion of the accumulated score value in the total weight of all score values is counted, and the slicing weight vector F is finally determined, where Fi represents the weight of the i-th network slice. The calculation process is shown in Equation (8).
4. Results and Simulations of Performance
4.1. Simulation Scenario
To verify the effectiveness of the network slicing strategy, we select 5G mobile communication base station layout data provided by local government of a coastal city in Southeast China as reference. We chose conducting the experiment in this coastal city because it is one of the first batch of 5G network coverage cities in China, and the construction of 5G base stations around the central area of this city has progressed well. Therefore, the construction of 5G mobile communication base stations in this city can fully meet the environment needed for our 5G slicing strategy experiment. Python is used to simulate physical network and NG-RAN base station. The layout data of base stations is shown in Figure 5.
Considering the complex topography and high-population density in the city center [25], the parameters of simulation experiments were set as seven types of slices, five modes of communication mobility, 20 NG-RAN base stations, and three types of users for testing.
The seven slice types set in Table 2 include three categories of eMBB slicing, one category of URLLC slicing, one category of mMTC slicing, and two categories of noise slicing, as well as five categories of network performance indicators, including guaranteed bandwidth (single user), maximum bandwidth (single user), guaranteed throughput (single slice), QoS priority, and maximum delay (single user). Because the eMBB scenario requires a large amount of bandwidth, the guaranteed bandwidth, maximum bandwidth, guaranteed throughput, and QoS priority are set to a higher level. The URLLC scenario has high-transmission delay and reliability, so the maximum delay is set to a small value, and the bandwidth demand and QoS priority can be relatively lowered. There are a large number of communication connection devices in mMTC scenario, so it is set as background service traffic to ensure effective transmission under large scale communication. Finally, the horizontal and vertical noise traffic are, respectively, set as the reference terms.
The weighted normalization matrix V can be obtained by calculating the network slicing data in Table 2 above through Equation (4). Then, the slicing weight vector F can be calculated according to Equation (8), and each item in vector F can finally be used as the weight of each network slicing type in simulation experiments, that is, the values of slicing weight in the last column in Table 2.
Five parameters of movement type are set to simulate user 5G data transmission in a complex urban environment. There are three types of pedestrian movement, which are stationary, free, and walking, as well as the driving and bus types with faster driving speeds, representing five speeds from low to high. The above five scenarios were simulated by normal distribution and random distribution, and the user weights were reasonably allocated. The five parameters of movement type in urban complex environment are shown in Table 3.
Table 4 shows the coordinates and coverage of the base station in the simulation experiments, and specifies its throughput and supported slice types. The number of the base stations reached 20, so the part of base station parameters is listed and shown in Table 4. As for the distribution of user information, stochastic distribution is adopted for modeling in horizontal and vertical axes and usage frequency. The specific parameters in client information are shown in Table 5.
4.2. Analysis of Simulation Scenario
Figure 6 shows the wireless network resource scheduling of seven types of services under 5G network slicing allocation when the simulation time is set to 600 s and the number of users is 1,000, 3,000, and 5,000. In the simulation system, the running time of the base station initialization test and the test time of the end phase are set to 30 s, so 0–30 s and 571–600 s are the initialization and finalization of the system, so only the average data in the effective time period (30–370 s) is used to analyze the impact of network slicing strategy on the overall performance of the base station.
(a)
(b)
(c)
According to the above description of 5G network slices, when the number of users increases, the actual throughput of the base station has an upper limit, leading to increased network load, decreased number of user connections, and increased blocking times. However, the network slicing strategy can effectively schedule the wireless network resources, which can greatly improve the resource utilization rate and reduce the frequency of base station handover. When the number of users is 1,000 and the running time is 600 s, Figure 7 shows the trend comparison of user connection success rate, total bandwidth occupancy, single slice bandwidth occupancy, single slice user occupancy, base station coverage rate, base station blocking rate, and base station hand over rate with or without network slicing strategy.
(a)
(b)
(c)
(d)
(e)
(f)
(g)
By analyzing the data in Figure 7 of system operation, the performance comparison of base stations with or without network slicing strategy is summarized, as shown in Table 6.
Through the analysis of the data in Table 6, it can be seen that after the network slicing strategy was adopted, the success rate of base station user connection during the test period remained at about 91%, and the connection efficiency was improved by nearly 29% compared with 64% when the network slicing strategy was not adopted. The total bandwidth occupancy of the base station fluctuates around 55.433 Gbps. The bandwidth occupancy rate of single slice and the number of users occupancy of single slice were stable at about 8% and 6.53, respectively. At the same time, the base station coverage rate is relatively stable, as high as 94%. When it is compared with the total bandwidth occupancy rate of 46.004 Gbps, the single slice bandwidth occupancy rate of 6%, the single slice user occupancy number of 4.64, and the base station coverage rate of 77% before the network slicing strategy was adopted, it can be seen that the network slicing strategy mainly improves the single slice connection number, coverage capacity and bandwidth performance of the base station greatly. However, in terms of blocking and handover, although the average rates of blocking and handover of base station have been reduced to 6.28% and 0.14%, compared with 19.41%, and 1.18% before the network slicing strategy is adopted, the performance is indeed significantly improved.
However, the stability still shows a certain degree of oscillation, especially in the base station handover, the fluctuation is relatively severe. Thus, we can speculate that complex urban environment in the future 5G network slicing technology, by taking a wide range of person traffic into account as large and unpredictable factors, user behaviors are more likely to lead to network performance bottlenecks of base station in blocking and handover. It is necessary to improve resource scheduling strategy for base station, to optimize the experience when users are actually using the network.
5. Conclusion
In this paper, we have designed a 5G slicing strategy for three typical application scenarios: eMBB, URLLC, and mMTC, which can provide a reference for the radio resource scheduling of city center base station by analyzing five indicators: occupied bandwidth, maximum bandwidth limit, occupied throughput, QoS class, and link tolerable delay.
To simulate physical network and NG-RAN base stations, we use 5G mobile communication base station layout data provided by the local government of a coastal city in Southeast China in the simulation experiments. The experimental results have shown that, compared with the case that the network slicing strategy is not adopted, when the number of users increases, the number of the blocking and handover of the base station decreases significantly in the case that the network slicing strategy is adopted, which can be reduced to about 6.28% and 0.14%. In addition, 91% of the user connections and 55.433 Gbps of bandwidth can also maintain a high value for a long time. At the same time, 94% base station coverage, 8% single slice bandwidth occupancy, and 6.53 single slice user occupancy are stable without significant fluctuations, which can greatly improve the overall utilization of 5G base station wireless resources. However, the network slicing strategy has a great impact on base station blocking and handover, and the fluctuation is relatively severe. In the future study, the corresponding performance optimization is also an interesting research point.
Data Availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Conflicts of Interest
The author declares that there is no conflicts of interest.
Acknowledgments
The authors acknowledge the Basic Scientific Research Project of Wenzhou (grant: G2020021) and Key Laboratory (Engineering Center) Construction Project of Wenzhou (grant: ZD202003).