In my previous article, "FAQs about 5G", I discussed about some of the most frequently asked questions about 5G. Even though this article was authored not long ago, I'm certain that the 5G development in China has achieved many significant breakthroughs since then. After taking the technical seminar "Technology Has Something to Say," I prepared this article to share some new ideas and best practices about 5G. You can take it as a continuation of my previous article, FAQs about 5G.
This article is split into three sections: understanding 5G based on statistics, understanding 5G from a technical perspective, and 5G products of cloud computing companies. 5G means high bandwidth, low latency, massive connectivity, and the power to drive the development of thousands of industries. It brings about challenges to the existing infrastructure, business architectures, business forms, and business models of companies, and new opportunities.
At the sixth World Internet Conference held in Wuzhen, Zhejiang, 5G was a hot topic discussed by the attendees. One CEO, used "SpeedTest" to test the speed of the 5G network at the conference venue. According to the following speed test results that I captured from Lei Jun's Weibo, the download speeds on the 5G network varies significantly from 81.5 Mbit/s to 787 Mbit/s. The speed difference is so large that it will affect the business performance of services deployed on this network, but even 81.5 Mbit/s, the lowest download speed shown on Lei Jun's Weibo, is much faster than many existing 4G networks. Whether it is slow or fast depends on how you define a weak network. After all, the lowest download speed in a 4G network could also be the highest in a 3G network.
If you take a closer look at the speed test results, you will find another metric in the upper-left corner: ping. All three test results have a high ping. In this respect, 5G does not have much advantage over 4G, and the ping is way higher than 1 ms, which is widely publicized by the media. Technically speaking, the ping depends on the testing server used by SpeedTest. However, the ping latencies shown in these three test results are still unsatisfactory.
Lei Jun's single-point speed testing results cannot fully reflect the current situation of 5G networks. The following testing results from the China Academy of Information and Communications Technology (CAICT) on 5G network performance in Beijing from November 2019 offer more ideas.
Let me explain some abbreviations used in the test results of CAICT. Call quality test (CQT) is a term initially used in call testing. A CQT is performed in multiple fixed locations, and the testing devices remain fixed throughout the test. A drive test (DT) is performed on moving vehicles on the road at a speed higher than 20 kilometers per hour. This test focuses on the performance of the 5G network when testing devices are moving. For this test, Daxing Airport and the building of the Beijing Organizing Committee for the 2022 Olympic Winter Games (BOCOG) were selected as the locations for CQT, and the Second Ring Road and Chang'an Street were selected for DT. In the chart on the left, you can see the 5G signal coverage at CQT locations is quite good. However, during the DT test, the lowest 5G camping ratio is only 62%. This result was obtained in the Tongzhou district, where 5G signals were absent in nearly 40% of the area during the test. According to my analysis, there is a high probability that 4G and 5G signals will coexist for a long time. In other words, we must ensure that our services can operate smoothly when the network switches back and forth between 4G and 5G. The chart on the right shows the average download speed of 5G networks in China in 2019, which is about 800 Mbit/s for CQT and 250 Mbit/s for DT. The overall result is not bad, considering the telecom operators have just received the 5G license. However, people may still feel a little disappointed because the previous publicity raised our expectations. Take a look at the following news:
When connected to the same standalone (SA) 5G private medical network from locations on the campus of the hospital and when connected from other regions in Henan province, the testing devices reported a consistent download speed of 918 Mbit/s and an upload speed of 180 Mbit/s. The upload speed of an SA 5G network is 50% higher than that of a non-standalone (NSA) 5G network. An average ping of 8 ms is reported by devices that are connected to this SA 5G network from locations on the campus, and 14 ms by devices that are connected from other regions in Henan province.
The preceding paragraph was part of the testing results of the SA 5G private medical network of the First Affiliated Hospital of Zhengzhou University. This test was performed on November 1, 2019, and the results were quite impressive. I wonder when these results can be achieved in general scenarios.
South Korea was the first country in the world to put 5G technology into commercial use. It officially started 5G services at the beginning of 2019. Their statistics can provide us with some references. First, look at some statistics from South Korea:
According to the statistics, it took South Korea nearly a year to increase the number of 5G subscribers to nearly 10% of the 4G subscribers, and 5G subscribers use two times more data than 4G subscribers. Currently, 5G services provided in South Korea are mainly HD videos. There are no statistics about the much-expected application of 5G in industry or IoT. In China, the average monthly data traffic of 4G subscribers in 2019 is 8.5 GB, in comparison to just 5.6 GB in 2018.
We can roughly take 5G device owners as 5G subscribers. Then, the ratio of the number of 5G subscribers compared to 4G subscribers in China would be close to South Korea. Note: China commercialized 5G technology more than half a year after South Korea. The average monthly data used by Chinese 4G subscribers is similar to South Korean 4G subscribers. The average monthly data usage of Chinese 4G subscribers is also increasing year by year. According to the 5G development journey of South Korea, there is a high probability that the data usage of Chinese 5G subscribers will triple. As my work is closely related to networks, I have to pay extra attention to network costs. Now, I am a little worried about whether the network bandwidth costs will triple too.
To make the 4G-to-5G upgrade easier for subscribers, Chinese telecom operators inherited their practices from the 3G-to-4G upgrade. To enjoy 5G services, Chinese subscribers only need to change their devices. They do not need to change their SIM/UIM cards, phone numbers, or data plans. Technically speaking, replacing 4G SIM/UIM cards with 5G ones is necessary to implement new security mechanisms. However, they gave up the approach of replacing SIM/UIM cards to encourage more subscribers to upgrade to 5G. In this approach, the only cost for the upgrade is the cost of their devices. How much does a 5G device cost? On December 10, 2019, new 5G phone for CNY 1,999 (USD 286). This is the first time a 5G phone was sold cheaper than CNY 2,000.
In the preceding figure, you can see two prices. Many people find a 5G phone for CNY 1,999 affordable. Some of you may find CNY 999 (USD 143) acceptable for a 5G industrial module when comparing it to an industrial module offered by Qualcomm. The latter costs about CNY 3,000 (USD 431). However, CNY 999 is far from appealing, considering that a 4G module costs less than CNY 100 (USD 14) on average. It will be very difficult to persuade tens of thousands of industrial enterprises to upgrade their 4G modules to 5G at this price. Therefore, it will take a long time for them to implement a 5G IoT.
Now, let's take a look at the 5G data plans offered by Chinese telecom operators. Even though you do not have to change your data plan to use 5G in China, most users will need more data based on the previous statistics. As a result, all three Chinese telecom operators released 5G data plans.
You can see in the preceding figure that these three telecom operators provide similar 5G data plans. Their cheapest data plan costs CNY 128 (USD 18) per month and provides 30 GB of data. According to the preceding statistics, 30 GB should be enough for most people. From the operators' perspective, the lowest price of CNY 128 is not profitable, but it is still a little high for low-income subscribers. Currently, the average monthly cost of 4G data plans is less than CNY 50 (USD 7) in China. Therefore, there is a high probability that the operators will offer discounts to reduce the monthly cost of 5G data plans in the future.
Most subscribers care about when 5G networks would cover areas where they live, travel, and work. Specifically, they want to know when seamless 5G coverage or the same level of coverage provided by 4G networks will be available. Essentially, this question is about the number of 5G cell towers that must be built in China. By the end of 2019, China had more than 5 million 4G cell towers, which is about half of the total number of 4G cell towers worldwide. We spent six years building them. 5G uses higher frequency bands than 4G but transfers information within a shorter distance. To achieve the same level of coverage as 4G networks, we need twice or three times the number of 4G cell towers; about 10 million 5G cell towers. Let's look at the following industry analysis report prepared by two security companies to get a basic idea about the construction period of 5G networks.
Here are some conclusions provided alongside this figure:
To sum up, a large number of target customers of Alibaba services may become 5G subscribers by the end of 2020, and we must adapt to the change as soon as possible.
The International Telecommunication Union (ITU) issued the International Mobile Telecommunications-2020 (IMT-2020 Standard) requirements for 5G networks and defined some network metrics as shown in the diagram on the left side of the following figure. These metrics are mapped to three well-known categories of 5G New Radio (NR) use cases defined by the 3rd Generation Partnership Project (3GPP): enhanced Mobile Broadband (eMBB), Ultra Reliable Low Latency Communications (URLLC), and massive Machine Type Communications (mMTC).
Look carefully at the metrics diagram on the left side of the preceding figure. You can see two shapes in it. The one in lighter green indicates IMT-Advanced requirements for 4G, and the one in darker green indicates IMT-2020 requirements for 5G. You can use this diagram to verify whether your business needs 5G. If all the network requirements of your business fall in the lighter green area, you only need 4G. If some of your requirements fall in the darker area, you need 5G. You may ask whether there are IMT-2000, IMT-2010, or IMT-2030. The answer is, there are IMT-2000 requirements for 3G and IMT-2030 requirements for 6G, but there is no IMT-2010. Then you may ask about the relationship between ITU and the generations of mobile systems. The IMT-2020 requirements were issued by ITU in 2015 for networks in 2020. Then 3GPP proposes 5G as its solution that meets the IMT-2020 requirements. Some of you may wonder what 3GPP is. In the telecommunications world, standards are fundamental to the business of operators and device manufacturers. Without unified standards, their networks and devices would not be able to communicate with those of others. 3GPP unites seven telecommunications standard development organizations (ARIB, ATIS, CCSA, ETSI, TSDSI, TTA, TTC) and provides their members with a stable environment to produce the Reports and Specifications that define 3GPP technologies. 3GPP specifications define the architecture, interfaces, and features of a communication system. The following figure shows the release control process of 3GPP for 5G.
Generally, communication devices are manufactured based on the corresponding specifications only when a release is frozen. Then, operators can purchase and deploy these devices. Therefore, the release freeze time is very important. As shown in the preceding figure, the freeze time for R15 and R16 is very important. R15 is the first SA release that can be commercialized and mainly focuses on eMBB. R16 is an SA release that adapts to multiple use cases and fully meets the IMT-2020 requirements. R16 is most important to us. It was completed on July 3, 2020, which marks the completion of the initial full 3GPP 5G system.
More 5G system enhancements are set to follow in R17. I will write about the details in another article.
Architecture is the first thing to consider for both services and networks. Knowing the architecture of a 5G network allows us to focus our research on the target problems from a broad view. Some of you may not be familiar with the architecture of cellular networks, but you may have some basic ideas about 4G. Let's look at the architectures of 4G and 5G networks to see what changes 5G brings to 4G.
In the following figure, the 4G architecture is shown on the left side, and the 5G architecture is shown on the right side. In the 4G architecture, the Packet Data Network Gateway User Plane (PGW-U) connects to the Internet to serve as the traffic egress. The Mobility Management Entity (MME) and Serving Gateway (SGW) connect to the 4G cell tower to serve as control units. It seems simple, but as the colors suggest, components with two or more colors, such as the MME serve two or more roles. The 4G architecture is characterized by fixed connections, features, and signal interaction paths. In other words, the 4G architecture is fixed.
In the 5G architecture, all components are shaded with only one color, indicating that each component serves only one role in a 5G network. This architecture is similar to the service-based architecture (SBA) of the IT industry. Most components are named in the form of "xxF", where F indicates function and xx indicates the function name. The point-to-point communication between components in the 4G architecture is replaced with bus-based communication. This architecture is the foundation of innovative technologies, such as network slicing and edge computing, which will be discussed later in this article.
The SBA depends on the network mode: non-standalone (NSA) or standalone (SA). 5G phone manufacturers must specify the network mode their phones support. China Mobile does not allow mobile phones that only support the NSA mode to access its 5G core network starting from 2020.
What is the difference between the NSA and SA network modes? As shown on the left side of the preceding figure, in an NSA network, the 5G cell tower is attached to an existing 4G cell tower. It leverages the 4G core network and has low construction costs. In an SA network, the 5G cell tower is connected to the 5G core network. The SA network provides all 5G features but has high construction costs. China Telecom, China Unicom, and China Mobile prefer the SA network mode. However, most of their 5G networks were built in the NSA mode in 2019 because the government's 5G policy rolled out too fast, and they didn't have enough time. The SBA is shown on the right side of the preceding figure, which is similar to the microservice architecture proposed by Alibaba. This shows how the information technology (IT) and communication technology (CT) industries are developing in the same direction.
What excites Telecom operators in the 5G era is how they can provide diversified services for more customers in different scenarios, not only network access services. Among the three major categories of 5G use cases I mentioned earlier, only eMBB is consumer-oriented. Each of the three categories covers many use cases in different industries. How does a single network serve all of these business requirements? The concept of network slicing is proposed to make sure the 5G network can meet your SLA requirements.
The preceding figure shows how network slicing works. You can see a sliced network as a highway that is divided into several lanes to meet the diversified business requirements. Network slicing involves parameter design, physical resource allocation, and process streamlining. It sounds simple but is a huge project to implement, especially when it involves a large number of users. The following figure shows a simple network slicing process.
First, the customer provides network service requirements (slice information.) Then, an operator (one of the three major telecom operators in China, or a secondary operator) incorporates the service requirements into the corresponding capabilities and enables these capabilities on a physical network. After that, the physical network generates the corresponding slices. Finally, the customer uses the network slices that it has subscribed to. As shown in this process, the customer must be clear about its network requirements, which include the network bandwidth and latency guarantee.
In the 5G era, operators provide network slicing and edge computing services. First of all, edge computing was created much earlier than 5G, and 5G is incomplete without edge computing. 5G accelerates the development of the edge computing industry. Why does 5G need edge computing? Well, it needs edge computing to reduce the air interface latency to less than 1 ms. Why isn't edge computing important for 4G? This is due to the difference between 4G and 5G architectures. As seen in the preceding architecture analysis, the 5G architecture contains a new component, User Plane Function (UPF). The physical location of this component must be determined flexibly to implement the proximity forwarding of business traffic and edge computing. The following figure shows an ideal edge computing-based 5G architecture.
The UPF node can be arranged at a location that is 50 km, 300 km, or 1000 km away from the nearest 5G cell tower. The operator provides edge computing resources to facilitate business deployment and provides services based on the proximity principle. Is edge computing merely about closer deployment?
The preceding figure shows a simple architecture of a non-CDN business. In this figure, the client is a mobile app, and the server is deployed in several cloud data centers. A centralized cloud-based business architecture allows developers to focus on the business logic without worrying about the underlying problems. In the case of edge computing, the architecture is changed.
The edge node shown in the preceding figure may comprise multiple levels. It is a huge challenge to unify the client, edge, and server and ensure they perform their duties properly. This is similar to the CDN architecture, but it is still different. If they are static resources, the only change in the architecture is that a new node is added. If they are dynamic resources or businesses, adding a new node creates a completely new architecture. They are more closely dependent on each other than nodes in a CDN architecture. What does it have to do with 5G? Edge computing seems to be irrelevant to 5G, except it can be used to determine the optimal location of the UPF node.
As shown in the preceding figure, a 5G network also has a Network Exposure Function (NEF) component. It is used to expose the network. I will share how to use this component to implement some interesting functions in another article.
We wish you a wonderful year with 5G!
Alibaba Clouder - July 19, 2018
Alibaba Clouder - January 19, 2021
XG Lab - January 13, 2021
Alibaba Clouder - September 27, 2020
Alibaba Clouder - March 25, 2019
Alibaba Clouder - December 2, 2020
Provides secure and reliable communication between devices and the IoT Platform which allows you to manage a large number of devices on a single IoT Platform.Learn More
Link IoT Edge allows for the management of millions of edge nodes by extending the capabilities of the cloud, thus providing users with services at the nearest location.Learn More
A cloud solution for smart technology providers to quickly build stable, cost-efficient, and reliable ubiquitous platformsLearn More
Deploy custom Alibaba Cloud solutions for business-critical scenarios with Quick Start templates.Learn More