5g 为什么需要边缘云_5g将云的力量带到了边缘

5g 为什么需要边缘云

by Peter McHale, Summer Associate, Darling Ventures

作者：Darling Ventures夏季合伙人Peter McHale

Big data management, SAAS, predictive analytics, and AI — these are just a few industries enabled by cloud computing. So much of the recent innovation we see in software has been unlocked by the cloud, you would be forgiven for thinking that everything is running on AWS or on another server farm. IoT, however, is a huge market of game-changing technology that has not been able to partake in cloud computing’s accessible economics and powerful computing capabilities. IoT applications often need an immediate reaction to what they sense, generate massive amounts of data to be processed, and have unique security constraints.

大数据管理，SAAS，预测分析和人工智能-这些只是云计算支持的少数几个行业。 我们在软件中看到的许多最新创新已经被云解锁，您认为一切都在AWS或另一个服务器场上运行就可以了。 但是，物联网是一个巨大的改变游戏规则的技术市场，它无法参与云计算的可访问经济性和强大的计算功能。 物联网应用通常需要对其感觉立即做出React，生成大量要处理的数据，并具有独特的安全性约束。

5G is about to change that. With its ultra-low latency, ability to cheaply send large amounts of data, and improved security, 5G will empower many IoT devices with the computational power of the cloud. When 5G powered IoT is able to utilize the capable AI algorithms that live in the cloud, we will see an IoT explosion of possibilities become reality. This shift offers great opportunity for companies working in the IoT space but also warrants a reexamination of their strategy and value proposition.

5G即将改变这种状况。 5G具有超低的延迟，廉价地发送大量数据的能力以及更高的安全性，将使许多物联网设备具有云计算能力。 当支持5G的IoT能够利用云中强大的AI算法时，我们将看到IoT爆炸的可能性变为现实。 这一转变为物联网领域的公司提供了巨大的机会，但也需要重新审视其战略和价值主张。

背景 (Background)

In the early 2000s, multiple startups, such as Marc Andreessen’s Opsware, popped up with a cloud computing product before most understood what the cloud was. Large companies such as Amazon and Microsoft later came out with offerings like AWS and Azure that made cloud computing as widespread as it is today. Outsourcing cloud computing to offerings such as AWS was attractive to companies for a number of reasons: 1) companies were able to have cheap upfront costs without having to buy expensive server hardware, 2) companies gained unprecedented reliability by shifting the responsibilities of maintaining servers to outside vendors, thus maximizing uptime and 3) companies were able to manage diverse web-traffic rates by paying as they went with unlimited accessibility to more computing resources. So why hasn’t IoT participated in these huge benefits?

在2000年代初期，马克·安德森(Marc Andreessen)的Opsware等多家初创公司突然涌现出了云计算产品，这才让大多数人了解云是什么。 像亚马逊和微软这样的大公司后来推出了像AWS和Azure这样的产品，它们使云计算像今天一样广泛地传播。 将云计算外包到AWS之类的产品对公司具有吸引力，其原因有很多：1)公司无需购买昂贵的服务器硬件即可获得便宜的前期成本； 2)公司通过将维护服务器的职责转移到外部供应商，从而最大程度地延长了正常运行时间； 3)公司能够通过无限制访问更多计算资源的方式付费来管理各种网络流量费率。 那么，为什么物联网没有参与这些巨大的利益呢？

IoT applications have been limited by three constraints that have kept them using computational and data storage resources at “the edge”, or close to where the data is created and the application is utilized. Many of these applications have unique security requirements that prevent them from sending their data to an external server. Other applications are constrained by not being able to cheaply send large amounts of data through the internet to a server. Finally, some applications need ultra-low latency to react quickly to data or input at the edge. 5G is about to meet these needs for many IoT applications and allow them to reap the benefits of the cloud. Now let’s explore this 5G IoT revolution.

物联网应用受到三个约束的限制，这使它们无法在“边缘”或接近创建数据和使用应用的位置使用计算和数据存储资源。 这些应用程序中的许多应用程序都具有独特的安全性要求，以防止它们将数据发送到外部服务器。 其他应用程序由于无法通过Internet廉价地将大量数据发送到服务器而受到限制。 最后，某些应用程序需要极低的延迟才能对边缘的数据或输入做出快速React。 5G即将满足许多物联网应用程序的这些需求，并使它们能够从云中受益。 现在让我们探索5G IoT革命。

迁移到云端 (Moving to the Cloud)

Alexandre Debiève on AlexandreDebiève在Unsplash上 Unsplash 拍摄

One area that is receiving plenty of attention from 5G media is VR, and for good reason. Currently, VR means buying high-priced hardware for an experience you likely aren’t using all the time because there isn’t a fast-enough means of receiving data that isn’t within a few feet of you. There’s already talk about using ultra fast latency to move this processing to the cloud, and bring about pay-as-you-go virtual reality as a service. 5G will bring this reality closer, but before we see that total paradigm shift, 5G is enabling VR to utilize cloud resources. BAD VR is an early such example, focusing on how VR can be used for visualizing data analytics. Imagine a world in which emergency phone operators update a back-end that lives in the cloud with data on a current emergency, while first responders wear augmented reality headsets that help them visualize all the data about the current incident, as well as any medical guidance to support their efforts once they arrive. All of this data would stream reliably and instantly via 5G when every second counts . This is the world that BAD VR is working to create.

VR受到5G媒体的广泛关注，这是有充分理由的。 当前，VR意味着购买昂贵的硬件，以获得您可能不会一直使用的体验，因为没有足够快的方法来接收不在您几英尺之内的数据。 已经有关于使用超快速延迟将处理转移到云中并带来即付即用的虚拟现实即服务的讨论。 5G将使这一现实更加接近，但是在我们看到整体模式转变之前，5G使VR能够利用云资源。 BAD VR就是这样的早期例子，专注于如何将VR用于可视化数据分析。 想象一下这样一个世界：紧急电话运营商使用当前紧急情况的数据来更新驻留在云中的后端，而第一响应者戴着增强现实耳机，以帮助他们可视化有关当前事件的所有数据以及任何医疗指导一旦他们到达就支持他们的努力。 每秒钟都可以通过5G可靠，即时地传输所有这些数据。 这就是BAD VR正在努力创造的世界。

While BAD VR is working to equip first responders navigating a city, others are focused on the city itself. Cities will become smarter and smarter via lightweight and cheap sensing hardware scattered around them. This hardware will communicate with the cloud, hosting large-scale optimizations and coordinating thousands of tasks using powerful AI. AI will then communicate with vehicles and pedestrians to create some remarkable mobility efficiencies. Haas Alert is an early player in this space. Their product enables first responders to communicate information to a “safety cloud” that then alerts motorists immediately of an accident so that they can slow down well in advance of seeing it. While there is some manual data input happening here, we will see more and more of these applications being automated.

虽然BAD VR致力于为急救人员配备导航城市的设备，但其他人则专注于城市本身。 通过散布在周围的轻巧廉价的传感硬件，城市将变得越来越智能。 该硬件将与云通信，托管大型优化并使用强大的AI协调数千个任务。 然后，AI将与车辆和行人通信，以创造出显着的移动效率。 Haas Alert是该领域的早期参与者。 他们的产品使急救人员能够将信息传达到“安全云”，然后立即向驾驶员发出事故警报，以便他们能够在看到事故之前放慢速度。 尽管这里发生了一些手动数据输入，但我们将看到越来越多的这些应用程序是自动化的。

Navilens is another early contender in the smart city space. Their app scans QR-like codes placed throughout a city to help people with visual impairments navigate through the city, or to help unimpaired pedestrians gain access to navigational information that is relevant to their exact location. This allows people to use the same technology outdoors or indoors (where GPS doesn’t work so well), potentially even while using a subway. They’re currently limited by the number of images scanned, however 5G’s data rates will greatly lift that constraint with access to cloud compute resources and will enable more seamless and exact navigational information.

Navilens是智慧城市领域的另一位早期竞争者。 他们的应用程序会扫描遍布整个城市的类似QR码，以帮助有视觉障碍的人在整个城市中导航，或帮助无障碍的行人获得与其确切位置相关的导航信息。 这样一来，即使在使用地铁的情况下，人们也可以在室外或室内(GPS无法正常工作)使用相同的技术。 它们目前受到扫描图像数量的限制，但是5G的数据速率将极大地限制对云计算资源的访问限制，并将启用更加无缝和准确的导航信息。

Smart manufacturing plants need quick responses to failing machinery or process failures. When 5G brings increased security and ultra-low latency, plants can send non-sensitive data to platforms and resource-intensive AI living in the cloud to perform analysis and then immediately respond at the plant. Petasense is a great example of this. A plant places a Petasense device on a machine that then sends accelerometer data to a platform in the cloud that uses AI. With the combined compute power of the cloud and the data from the IoT sensor in the plant, an AI can detect a failing machine and send a signal to the plant to shut down and schedule maintenance. Augury, which Qualcomm recently invested $8 million in its Series C, offers a similar product and aims to catalyze a wave of valuable IoT manufacturing technologies that are possible with 5G.

智能制造工厂需要对出现故障的机械或过程故障做出快速响应。 当5G带来更高的安全性和超低延迟时，工厂可以将非敏感数据发送到生活在云中的平台和资源密集型AI进行分析，然后立即在工厂进行响应。 Petasense是一个很好的例子。 工厂将Petasense设备放置在机器上，然后将加速度计数据发送到使用AI的云平台。 借助云的计算能力和工厂中IoT传感器的数据的组合，AI可以检测出故障的机器并将信号发送给工厂以关闭并安排维护。 高通最近在其C系列中投资了800万美元，提供了类似的产品，旨在催生一波有价值的物联网制造技术浪潮，这可能是5G可能的。

5G物联网的局限性 (Limitations of 5G Powered IoT)

Photo by PAUL SMITH on Unsplash

PAUL SMITH在 Unsplash上的 照片

The IoT landscape is about to go through a 5G-powered revolution, but not everything will change. Some IoT applications will still have to remain at the edge in this new world. What applications will remain at the edge? Throughout this piece, I referred to costs of communicating large amounts of data, unique security needs, and ultra-fast speed requirements as the reasons IoT hasn’t made its way to the cloud. Though 5G will greatly tackle these needs, it will still fall short for some applications.

物联网领域将经历一场由5G推动的革命，但并非一切都会改变。 在这个新世界中，某些物联网应用仍将必须保持优势。 哪些应用程序将保留在边缘？ 在整篇文章中，我都提到了通信大量数据的成本，独特的安全需求和超高速需求，这是物联网尚未进入云的原因。 尽管5G将极大地满足这些需求，但对于某些应用程序来说仍将不足。

Maximum Security Requirements

最高安全要求

Though 5G will offer incredible security via virtualization and network slicing, this security in communication doesn’t make the cloud servers your data is stored in any more or less secure, and so many applications will still opt to stay at the edge for higher security guarantees. 5G might still be utilized in these IoT devices, but as a means of getting data from distanced devices to a local server rather than to the cloud. Some applications that might fit into this bucket include industrial control systems and plant automation, such as Veo Robotics ‘ product which tracks human movement in order to allow safe collaborative work with automated machines. We might also see some hybrid approaches here to get the benefits of the cloud while keeping most of the security of hosting data at the edge, such as storing data on local servers, but then allowing a platform for operational management running in the cloud to access the local data.

尽管5G将通过虚拟化和网络切片提供令人难以置信的安全性，但是这种通信安全性并不能使您的数据存储在云服务器上或多或少地安全，因此许多应用程序仍会选择留在边缘以获得更高的安全性保证。 。 5G可能仍会在这些IoT设备中使用，但它是一种将数据从远程设备传输到本地服务器而不是云的方法。 可能适合该存储桶的某些应用包括工业控制系统和工厂自动化，例如Veo Robotics的产品，该产品跟踪人类的活动，以便与自动化机器进行安全的协作。 我们可能还会在这里看到一些混合方法，以获得云的好处，同时将托管数据的大部分安全性保持在边缘，例如将数据存储在本地服务器上，然后允许运行在云中的运营管理平台访问本地数据。

Still Too Much Data

数据仍然太多

The cloud can offer more cost effective computation, but there’s still a cost to get your data to the cloud server. While 5G is expected to decrease this cost, there still exists a tradeoff, and so some use cases will still find it worthwhile to keep the data processing at the edge. A good example of this is automated security systems. If an application consumes gigabytes of video data in a day, but all it needs to do with it is check if anything has changed frame-to-frame, then the engineer might keep this simple computation close to the camera and only send up data when something has changed, like Reolink ‘s Argus 2 camera does.

云计算可以提供更具成本效益的计算，但是将您的数据传输到云服务器仍然存在成本。 虽然预计5G可以降低成本，但仍然存在折衷，因此某些用例仍然认为值得在边缘进行数据处理。 自动化安全系统就是一个很好的例子。 如果应用程序一天要消耗数十亿字节的视频数据，但要做的只是检查帧间是否发生了任何变化，那么工程师可能会将这种简单的计算保持在摄像机附近，并且仅在以下情况下才发送数据：发生了一些变化，例如Reolink的Argus 2相机。

Optimally Minimal Latency

最佳最小延迟

While 5G has lightning fast latency characteristics (theoretically as fast as 1–4 ms), it will still add that time, plus the time to get a response from a server, to any process that relies on it to run an application in the cloud. In some settings, the value of every millisecond is greater than the benefits gained from using cloud computing. A great example of an application that will not leave the edge for this reason is the AI for autonomous vehicles, like that which May Mobility is creating. When you write a feature that will enable a vehicle to detect a pedestrian or a car, you will shave off every millisecond you can from the sensor making the measurement to the vehicle hitting the brakes. At the end of the day, you don’t want the reason a car hit someone to be that a cloud server, for whatever reason, took too long to respond.

尽管5G具有闪电般的快速延迟特性(理论上最快为1-4毫秒)，但它仍然会增加该时间以及从服务器获得响应的时间，并依赖该时间来在云中运行应用程序的任何进程。 在某些情况下，每毫秒的值大于使用云计算所获得的收益。 出于这一原因而不会落后的应用程序的一个很好的例子是自动驾驶汽车的AI，例如May Mobility正在创建的AI。 当您编写使车辆能够检测到行人或汽车的功能时，您会从进行测量的传感器到车辆踩下刹车的每一毫秒内剃光。 归根结底，您不希望汽车撞到某人的原因是云服务器(无论出于何种原因)花费了太长时间做出响应。

物联网公司在5G革命期间取得成功的策略 (Strategies for IoT Companies to Succeed During the 5G Revolution)

So change is coming, for better or worse, and for most it will be for the better! But if you are a company working in the IoT space and are concerned about 5G being a disruptive force, what can you do about it? I propose two strategies, and since this shakeup is around the future of product in IoT, the strategies are centered around product.

因此，无论好坏，变革都将到来，对于大多数人来说，它将变得更好！ 但是，如果您是一家从事物联网领域工作的公司，并且担心5G是一种破坏性力量，那么您将如何做？ 我提出了两种策略，并且由于这种变革是围绕物联网产品的未来发展的，因此这些策略以产品为中心。

First of all, as noted earlier, not all of IoT will be moving to the cloud. If your product is focused on applications that will stay at the edge, then you will be in a strong position for the next decade as your customers remain largely unaffected. One great example of this is one of our portfolio companies, Foghorn. Foghorn specializes in applying intelligent algorithms at the edge for industrial processes, and with the type of data they work with, many of their customers’ applications are going to be staying at the edge. First of all, the data can be sensitive, involving videos of manufacturing plant workers used to detect whether safety equipment is being appropriately worn and data on plant-specific operations that contain valuable IP for the company. Secondly, much of the data is massive, consisting of video streams, and would cost more to stream to a cloud server than it would to process locally with Foghorn’s edge-optimized AI. While many of their applications will likely remain at the edge, Foghorn is also well-positioned to take advantage of 5G for uses such as ultra-low latency automation.

首先，如前所述，并非所有的物联网都将迁移到云中。 如果您的产品专注于将保持优势的应用程序，那么在未来十年中您将处于优势地位，因为您的客户在很大程度上不会受到影响。 一个很好的例子就是我们的投资组合公司之一Foghorn 。 Foghorn专门在工业流程的边缘应用智能算法，并且由于使用的数据类型不同，许多客户的应用程序仍将停留在边缘。 首先，数据可能很敏感，涉及用于检测安全设备是否被适当磨损的制造工厂工人的视频，以及包含该公司宝贵IP的工厂特定操作的数据。 其次，许多数据是巨大的，由视频流组成，流向云服务器要比使用Foghorn的边缘优化AI在本地处理要花费更多。 尽管他们的许多应用程序可能仍处于边缘状态，但Foghorn的定位也很不错，可以利用5G进行超低延迟自动化等用途。

The second strategy is to focus on creating value in the IoT space independent of where the data is hosted or where the computation is happening. Anylog, a new startup, does exactly this. Their product allows you to navigate a decentralized IoT graph of data storage and computational resources, wherever the different hardware may be, as a singular centralized database. This product does not care if the data is stored in the cloud, at the edge, or a mix of the two. It also doesn’t care if the hardware for the computation you perform on the data is near the edge or in the cloud. It allows you to abstract all of that detail away, and parse the data as if it were collected in a single place. This kind of offering will keep companies valuable after 5G shifts many IoT applications to the cloud. This product strategy can even provide an advantage to companies that decide to shift from the edge to the cloud, as these companies would parse their data the same way before and after the shift.

第二种策略是专注于在IoT空间中创造价值，而与托管数据的位置或进行计算的位置无关。 新启动的Anylog正是这样做的。 他们的产品使您可以将数据存储和计算资源的分散式IoT图导航为单个集中式数据库，而无论硬件在哪里。 该产品不关心数据是存储在云中，在边缘还是在两者的混合中。 它也不在乎您对数据执行计算的硬件是在边缘还是在云端。 它使您可以抽象所有这些细节，并像将其收集在单个位置一样分析数据。 在5G将许多IoT应用程序转移到云后，这种产品将使公司保持价值。 这种产品策略甚至可以为那些决定从边缘迁移到云的公司提供优势，因为这些公司在迁移之前和之后都将以相同的方式解析其数据。

Hopefully throughout this piece you’ve noticed a common thread about what 5G and the movement from the edge to the cloud will do for IoT. It’s not just about existing applications moving their data and compute from the edge to the cloud, it’s also about entirely new possibilities. With the constraints that 5G will relax for IoT, there will soon be an IoT explosion with new applications and new devices popping up everywhere to collect data or communicate to the edge from an AI running in the cloud. Somewhat ironically, as this centralization of computing resources to the cloud decreases costs, offers an incredibly robust back-end, and makes scaling trivial for IoT, it will cause a decentralization revolution. In the coming decade, countless IoT products will flourish in a beautiful symbiosis between lightweight devices at the edge, cloud computing resources in the back-end, and 5G in the middle to communicate between the two.

希望在整个文章中，您已经注意到有关5G以及从边缘到云的移动将为IoT做些什么的共同话题。 这不仅涉及现有应用程序将其数据和计算从边缘迁移到云，还涉及全新的可能性。 鉴于5G放宽物联网的限制，不久将出现物联网爆炸式增长，新应用程序和新设备随处可见，以通过云中运行的AI收集数据或与边缘进行通信。 具有讽刺意味的是，由于将计算资源集中到云中可以降低成本，提供令人难以置信的强大后端，并使物联网的扩展变得微不足道，这将引发分散化革命。 在未来十年中，无数的物联网产品将在边缘的轻量级设备，后端的云计算资源以及中间的5G之间实现美好的共生，从而在两者之间进行通信。

Originally published at https://medium.com on August 13, 2020.

最初于 2020年8月13日 发布在 https://medium.com 。

翻译自: https://medium.com/@DarlingVentures/5g-brings-the-power-of-cloud-to-the-edge-bdaad4b658a0

5g 为什么需要边缘云