由DCD中国授权DKV编译并发表于DeepKnowledge微信公众号。

Peter Judge

Peter Judge是DCD全球编辑，他在英国从事科技B2B出版有多年工作经验，曾在Ziff-Davis, ZDNet, IDG 和Reed公司任职，专注于网络、安全、移动和云技术领域。

5G将成为承载边缘应用的关键，不过最大的边缘应用可能就是5G网络本身。

任何有关边缘的话题都绕不过5G，5G是下一代移动数据网络并将在未来几年中部署。5G将提供更高速的数据传输率，这是我们所听到宣传中预期边缘应用所必须的。

5G网络经过前几年的准备后，2019年4-5月间开始出现在大众视野。SK Telecom和其他运营商在韩国，Verizon和AT&T在美国，EE在英国也都推出了5G。其他国家和网络运营商也将快速跟进，也许在你看到这篇文章时又有更新。

此前的试验方都称在技术方面实现了预期目标。TechRadar在美国达到了1Gbps的测试速率，约500Mbps，明显比4G网络快很多。

全新之处

与前几代一样，5G也由运营商实现核心网络升级，包括从终端用户和设备到核心网络间的无线协议。新的无线技术被称为5G NR(新空口)，提供了高达到20Gbps的速率，不过该速度只是理论值，在这点上与无人能从今天的LTE网络中获得2.4Gbps峰值速度如出一辙。

5G NR在两个频段工作，一个是低于6GHz(通常3.5GHz)，另一个是30-300GHz的毫米波波段mmWave。低频波段与现有网络类似，目前广泛采用，是在4G的基础上进行了持续改善。毫米波mmWave的部署方式令人期待，可能提供更快的速度，要求更窄的频段和部署更多基站。

与前几代一样，技术将不断演进，以“non-standalone”（非独立组网）模式与现有核心网络共享。

媒体不断追问5G用户手机、热点、路由器和其他外设的进展。数字化基础设施产业也想了解：5G能为边缘应用如物联网IoT和无人驾驶带来哪些便利？

对物联网IoT而言，重要的一点是支持物联网传感器所需机器间通信的设备，在2025年将预计达到750亿美元的规模。支持低功耗广域(LPWA)网的标准，包括NB-IoT(窄带物联网)和LTE-M(LTE机器型通信)。

这些在5G的宣传中有些别扭，因为他们在此前针对4G标准中已经被定义过，现在突出相对较低速率，而此前灌输给客户时却是最大亮点。

物联网虽然需要数以亿计的传感器，但是大多数并不需要高的带宽。这些传感器主要是核对温度读数和其他数据等小的数据包，通常需要较长的电池供电时间，信号覆盖也广泛。5G可以实现前者，而在功耗和覆盖方面仍需要时间改善。在今后相对长的时间内，物联网方面4G完全够用。

有些人希望自动驾驶汽车是5G的“独家利器”，数据中心物联网厂商Wave2Wave的Duncan表示，如果要实现完全意义上的自动驾驶处理过程成本上极其昂贵。相反，他们需要从外部态势感知后进行毫秒级以下的响应，就能避免行人等障碍。

虽然5G会有时延，有可能丢失信号，这就需要在无人驾驶车辆中压缩必需的预置功能，从依赖5G中有效解脱出来。

与此矛盾的是，真正需要5G网络的应用程序也许正是此前提到的用户人群。他们产生与此前同样类型的数据，不过人们处理起来更快速了，这也带来了变化。

“有能源利用和延迟的制约，我们无法在集中化的数据中心基础设施处寻求解决方案”，Ellis在最近一次DCD网络直播中如此说道。一个用户需要1Gbps的数据传输就可以占用一个蜂窝基站的全部可用回程链路。

敷设到基站铁塔的光缆也许会解决这一瓶颈，成本却居高不下。目前来说，网络必须解决把数据传输到靠近客户端。Ellis补充说：“这意味着建设这些网络必须面临的巨大挑战，因为无从知道数据何时何地流向何方”。

最终，5G网络承载的最大边缘应用，也许就是5G网络本身，及其使用人群。

英文原文

Searching for the Edge on 5G

Peter Judge

Peter Judge is the global editor at DatacenterDynamics.，He has been involved with tech B2B publishing in the UK for many years, working at Ziff-Davis, ZDNet, IDG and Reed. His main interests are networking, security, mobility and cloud.

5G is supposed to be a key part of delivering Edge applications. But the biggest Edge application could be the 5G network itself.

Any talk of Edge inevitably comes round to 5G, the next-generation mobile data network, that is due to be deployed over the next couple of years. 5G will offer increased data rates, that we are told will be needed by predicted Edge applications.

After years of preparations, 5G networks began to appear in April and May 2019. SK Telecom and others are offering it South Korea, Verizon and AT&T have launched in the US and EE has 5G in the UK. Other countries and networks will follow swiftly - possibly by the time you read these words.

Early testers say the technology is living up to expectations. TechRadar clocked speeds of more than 1Gbps in the US, and around 500Mbps in the US, which is much faster than 4G networks.

The new New

Like previous generations, 5G upgrades the core networks run by the operators, and also the radio protocol by which they communicate to end-users and devices. The new radio technology, helpfully called 5G NR (for “new radio”) could offer speeds up to 20Gbps, but that speed will be a largely theoretical one - much like the 2.4Gbps top speed that none of us get from today’s LTE networks.

5G NR operates in two frequency bands, one below 6GHz (normally 3.5GHz), and the other is in the mmWave band, between 30-300GHz. The lower frequency band most like existing networks, is most commonly deployed now, and offers a moderate improvement over 4G. The mmWave option is deployments more exotic and likely to give faster speeds, at the cost of a shorter range, which will require many more base stations to be deployed.

Again, like previous generations, the technology will come in gradually, using “non-standalone” mode to share the existing core network.

The media is asking about 5G handsets, hotspots, routers and other paraphernalia used by humans. The digital infrastructure industry will want to know: what can 5G do for Edge applications, like the IoT and autonomous vehicles?

For the Internet of Things, the important thing will be equipment to support machine-to-machine communications for the IoT’s sensors - 75 billion of which are predicted to be hooked up by 2025. The standards support low power wide area (LPWA) networks, with options including NB-IoT (narrowband IOT) and LTE-M (LTE-machine type communication).

But these sit somewhat strangely in the 5G hype, as they were already defined for the previous 4G standards, and they deliberately use a lower speed than is being pushed as the major benefit for end-users.

The fact is that the IoT may have billions of sensors, but they mostly don’t need high bandwidth. They tick along handing over small packets of temperature readings or other data, and mostly need a long battery life and good coverage. 5G can potentially offer the first, but it will be some while before it approaches the second. For a long time to come, the IoT will find 4G completely adequate.

Some hope that autonomous cars could be the “killer app” for 5G. Duncan Ellis, of data center IoT firm Wave2Wave says it would be too expensive for them to carry enough processing to be fully autonomous. Instead, they will need sub-millisecond responses from external pattern recognition to avoid obstacles such as pedestrians.

But even 5G includes delays, and the possibility of lost signals, so there’s a strong strand in autonomous vehicle work trying to compress more of what is needed into the vehicle, effectively pulling back from relying on 5G.

Paradoxically, the application which will really stress 5G networks may be those human users we talked about earlier. They are pumping the same kind of data they always have, but they are doing it faster than ever before, and that may change things.

“With constraints like energy utilization and latency, we can’t keep going back to a centralized data center infrastructure,” said Ellis in a recent DCD webcast. One user taking 1Gbps of data would effectively take the entire backhaul available in one of today’s cells.

Fiber to the cell towers might ease that bottleneck, but that will be expensive. For now, the network will have to get better at moving the data closer to users. “This represents a huge challenge for those building networks, because you have no way of knowing at any time where the data flow is going to,” says Ellis.

In the end, the biggest Edge application on the 5G network, could be the 5G network itself, and its human users.

公众号声明：

本文并非原文官方认可的中文版本，仅供国内读者学习参考，不得用于任何商业用途，文章内容请以英文原版为准。中文版未经公众号DeepKnowledge书面授权，请勿转载。

关于公众号DeepKnowledge：

中立、系统、深入的数据中心专业知识公众号。

关于Deep Knowledge Volunteer（DKV）计划：

目前DKV创始成员已经满额，DKV正开放接受DKV精英成员加入，相关DKV精英成员计划加入细节请发邮件至dkv@dcinfra.com咨询。