Popular talk
Here to talk about a Tongsudejiang edge computing.
Why popular to say it, do not be afraid if popular, you did not understand. New things out of time, often require a process of acceptance and understanding. Like just came out before the Internet, many people do not know the Internet, so you have to slowly science, so that we slowly come to accept and understand it. Who now explain what it is the Internet.
The edge of the calculations for some time, but with the development of things, the concept of edge computing is becoming popular. We look at some popular non-introduce the concept of edge computing:
Edge computation, is a distributed computing architecture. In this architecture, the application operation, and data services, by the network central node, moved to the edge node on the network logic to handle.
Or, the edge of the original operation is handled entirely by the central node large service to be decomposed into smaller and more manageable portions, to the edge node to the dispersion process. The edge node is closer to the user terminal device, and to speed up processing of data transmission rate, to reduce delay.
Xiao Bian explained above excerpt is from the network edge computing for some articles. Basically the whole explanation jargon, most reading this passage, what is still not clear edge computing.
First, small series is not first give a good example. For example, there is a APP, APP users when using this, it will gather information about the user, such as collecting the user's age, gender, phone number, address location, search history, and so forth, and collect this information mainly better analyze the user's behavior and things of interest, such as interested in what car, house, books, food and so on. Then a more accurate delivery of its content and advertising.
This is a very common feature, but it is such a function, and how it linked to edge computing. Before edge computing, cloud computing is a.
If you are using cloud computing, this APP behavior is this:
After APP collected information, all the basic information uploaded to the server, and then by the server to perform the algorithm to calculate and identify user interests and may even figure out the user's spending power. According to the server can then extrapolated the results for the user to access their content and advertising interest.
If you are using edge computing, this APP behavior is this:
After APP collection of information, do not upload to the server. Then calculate and identify the user's interests and hobbies from APP yourself, you can also calculate the user's spending power, which is calculated functionality of the server, directly done by APP. The server then only need to ask APP, which the user is likely to be millions of annual salary, which is a single user. APP only need to tell the server that the user handsome, but also single, loves to travel, write poetry, can put their blind date beauty content.
In this way, the whole process was not involved in the calculation server, and the server was not involved in gathering information. Because this information is collected and calculated in APP itself, and not to upload, so it does not involve information gathering.
And, this is the edge of computing.
That is part of the server as previously calculated, is now replaced by direct calculation information collection device, and then the result of the calculation, outputs directly to the server. Server as long as the result does not need to process the data.
Edge computing, plainly, is (server) cloud computing bother to forget, the data on this point, you're in data collection, what of their own way, what are thrown into the server to count, very tired. As a result, the edge is calculated so came. Edge computing What are the advantages of it?
1, tun
Edge distributed computing and properties near the end of equipment destined advantage of real-time processing, so it can better support local business and the implementation of real-time processing.
2, is not simple and crude high efficiency
Doorstep things do not bother far away cloud computing, and edge computing for data terminal equipment to filter and analyze direct, while saving energy efficiency is higher.
3, worry and effort provincial traffic
Computing edge data explosion pressure and slow down the traffic network, the edge node used for data processing, data traffic from the device to reduce the cloud.
4, intelligent and more energy-efficient
AI + edge computation than calculating an edge in combination calculation, intelligent characteristics significantly, further calculating a combined edge cloud + attack, alone only 39% of the cost of the cloud.
Since such a cow edge computing, cloud computing can replace it?
Although the future will be more and more basic computing tasks to be accomplished edge, but this can only represent the edge of the device where the device will be more sensitive, but can not directly say that these tasks and clouds nothing to them is a let each other more perfect existence.
Edge computing and cloud computing collaboration with each other, they exist to optimize complement each other, enabling the industry together digital transformation. Cloud computing is a co-ordinator, which is responsible for large data analysis of long-period data, it can be periodic maintenance, and other operational areas of business decisions. Analysis focused on real-time, short-period data edge computation, better support local service process performed timely. Computing devices near the side edge, but also to contribute to the data acquisition cloud, cloud applications supporting large data analysis, but also by cloud sent to the edge of the large output data analysis business rules, and to perform the optimization process.
Calculating an edge (Edge computing) is a relative terms the cloud, it refers to collect and analyze the behavior of the local data generating device and the network close to the data generated. Distributed computing also called the edge of the cloud, mist, or fourth-generation computing data center.
Official speaking
Probably a lot of people have this experience: careless hand was on fire or open water hot, people will immediately remove his hand, this reaction is self-organizing conditions for human reflex response. We assume that, if our hands were burned or Shuitang reaction by our brains make decisions based on the information collection, and then take action, that would be what kind of scene?
Suppose we put people reflexively labeled edge computing, the human brain's response marked as cloud computing, then we can understand the difference between shallow and deep edge computing and cloud computing.
1. What is the edge computing
Calculating an edge (Edge computing) contrasting with the cloud, it refers to collect and analyze the behavior data generating device near the local network and the data generated, rather than having to transmit data to the cloud computing resources centralized processing . Distributed computing also called the edge of the cloud, mist, or fourth-generation computing data center.
Edge computing first emerged on the WAN network via virtual network services. Was originally developed by a platform to drive to adapt to the user's habits cloud computing, which is the origin of fog computing concept Cisco (Cisco) in 2011 was raised. With the advent of new edge computing power, edge computing is no longer need to build a centralized data center, the ability to create a large-scale distributed nodes potentially thousands of applications available.
II. Why do we need to calculate the edge
Gartner predicts that by 2020 up to 250 million smart devices will connect to the Internet, so much the device will generate 50 trillion GB of data around the world, which is quite more than five times the amount of global data in 2015. If all the data generated by these devices are transmitted to the cloud, network bandwidth, network traffic cost control, cloud storage capacity is a huge challenge. At the same time, some applications require a timely response, such as a failure of plant machinery and equipment forecast, the delay means that loss. In addition some of the edge device also relates to personal privacy and security. To meet the challenges of things in the scene of massive data transmission, storage and cloud computing capabilities, the leading cloud computing vendors have introduced edge computing products. The part of the data analysis functions into a nearby scenarios (the terminal or gateway) to achieve, intelligent services offered to meet near this industry in agile digital connection, real-time business data optimization, application intelligence, security and privacy protection, etc., the key requirements.
Edge computing Origin
Edge computing is a concept before the rise in recent years, its appearance is derived from cloud computing in the practical application of deficiencies:
Case 1: The intelligent manufacturing build factories, there will be a large number of intelligent terminals and network access equipment by industry, businesses need to calculate and handle increasingly large daily business data. At the same time, a large number of scenarios require real-time processing industry, the need for real-time response in milliseconds. Due to limitations of the network, cloud computing architecture is difficult to achieve real-time response. (Ie accident delay)
Case 2: unmanned vehicles need to respond to the environment in high-speed mobile state, so the response time is an extremely important indicator. Assuming that vehicle speed is 65 mph, emergency brake response time even if only for a few milliseconds slower, car emergency braking distance will be more of a few feet, which is perhaps the difference between an accident and the accident did not happen. (Delay that is life)
Case 3: The number of sensors, automated oilfield production data acquisition, if each sensor is coupled to the cloud transmission, the mass of data to the network tremendous pressure. (Ie, massive congestion)
Case 4: If your home is air-conditioning intelligent control, but relying on cloud computing. But you do not have power at home, but off the network, then how do? The cloud can not be controlled, even though you and the sweat, air conditioning also furnishings, this would not be very embarrassing? Edge computing solutions under the control of a network that is not the case. (No no network service)
1, cloud and calculating the difference between the edge
2, several traits edge computing
Distributed computing and low latency
Real-time calculation of an edge focus, short cycle analysis data can be better support local traffic and real-time intelligent processing performed
higher efficiency
由于边缘计算距离用户更近，在边缘节点处实现了对数据的过滤和分析，因此效率更高
更加智能化
AI+边缘计算的组合出击让边缘计算不止于计算，更多了一份智能化
更加节能
云计算和边缘计算结合，成本只有单独使用云计算的39%
缓解流量压力
在进行云端传输时通过边缘节点进行一部分简单数据处理，进而能够设备响应时间，减少从设备到云端的数据流量
三. 技术进步为布署边缘计算提供了可能
在物联网场景下，每个智能设备都会产生大量的数据，传输如此海量的数据从本地到云端，则需要消耗大量的网络带宽。为了加快服务和计算处理数据的时间，将计算从云端移向采集数据的边缘节点则是必然之选。
如上表所示，在30年前，计算通常发生在资源集中的大型机上。而20年前，随着PC的发展，C/S架构变得流行，任务处理变成分布式模型，客户端处理业务逻辑，数据库存储和交换数据。又经过10年的发展，为了提升用户体验、提供更敏捷的软件升级和改进，B/S架构占据主流，业务处理和存储又集中到了云端完成。现今，随着连入云端的智能设备越来越多、数据量越来越大，而且智能设备芯片的运算能力越来越强，这为使用边缘节点完成对初始数据的处理和分析便提供了必要的条件。
四. 怎么布署边缘计算
在物联网场景下，每个智能设备都会产生大量的数据，传输如此海量的数据从本地到云端，则需要消耗大量的网络带宽。为了加快服务和计算处理数据的时间，将计算从云端移向采集数据的边缘节点则是必然之选。其实，在大数据场景下，将计算部署到靠近数据的节点早有先例。Hadoop中的MapReduce就是通过将mapper和reducer部署到数据存储的节点，从而高效的处理HDFS中存放的海量数据。
边缘计算环境是构成物联网生态系统的诸多元素的一个子集，它剔除了管理、安全和分析功能。边缘计算是联接物理世界和虚拟世界的一道“桥梁”。
1、设备域：边缘计算在这一层，可以对感知的信息直接进行计算处理。比如在制造领域，可以对设备进行适时监控，能够实现预防性维护；在视频采集、音频采集中直接部署智能鉴别的能力;又或者像手机一样，能够由语音输入直接转换成文字输出。
2、网络域：通过部署计算能力，实现各网络协议的自动转换，对数据格式进行标准化处理。要解决物理网中数据异构的问题，就需要在网络域中部署边缘计算，以实现数据格式的标准化和数据传递的标准化(例如将所有的感知数据都换算成MQTT类型数据，并通过HTTP方式传递)。同时，网络域的边缘计算，还能对“融合网络”进行智能化管理，实现网络的冗余，保证网络的安全，并可进一步参与网络的优化工作。
3、数据域：边缘计算，使得数据管理更智能、存储方式更灵活。首先，边缘计算可以对数据的完整性和一致性进行分析，并进行数据清洗工作，消灭系统中的“脏”数据。其次，边缘计算可以对计算和存储能力、以及系统负载进行动态地部署。最后，边缘计算还能和云端计算保持高效协同、合理分担运算任务。
4、应用域：边缘计算提供属地化的业务逻辑和应用智能。它使得应用具有灵便、快速反应的能力，并在离线的情况下(和云端失去联系时)，仍能够独立地提供本地化的应用服务。
五. 边缘计算典型应用场景
边缘网络基本上由终端设备（例如移动手机、智能物品等等）、边缘设备（例如边界路由器、机顶盒、网桥、基站、无线接入点等等）、边缘服务器等构成。这些组件可以具有必要的性能，支持边缘计算。作为一种本地化的计算模式，边缘计算提供了对于计算服务需求更快的响应速度，通常情况下不将大量的原始数据发送云网络。然而，总体来说，边缘计算不需要会主动协助 IaaS,、PaaS、 SaaS和其他云服务，更多地专注于终端设备端。
边缘计算的概念是因工业制造之因而起。在工业领域，云端固然必不可少，但是仍需要边缘与云端的协同工作。单点故障在工业级应用场景中是绝对不能被接受的，因此除了中心云的统一控制外，工业现场的系统也必须具备一定的活力，能够自主判断并解决问题。边缘计算可以更便捷的处理工厂设备产生的海量数据，及时检测异常情况，更好的实现预测性监控，提升工厂运行效率的同时也能预防设备故障问题。
除工业制造之外，边缘计算在物联网时代不断增长的数据催生了对边缘计算的需求，下图是边缘计算的典型应用场景：
1、工业制造
边缘计算可以更便捷的处理工厂设备产生的海量数据，及时检测异常情况，更好的实现预测性监控，提升工厂运行效率的同时也能预防设备故障问题。
2、安全监控、ARVR
边缘计算提供快速、高效、精准的实时响应，将驱动安防行业人工智能应用迈入全新层次。
3、智能交通
智能交通信号灯可以根据路上车流的情况动态的调整信号灯的颜色，提高交通流畅度，减少拥堵，还可以应用于紧急情况，例如：信号灯可以为紧急情况开辟出一条绿色通道。
4、自动驾驶
自动驾驶在躲避障碍物的过程，若按照先上传云端、分析处理、再返回设备的模式，将造成信号传输的延迟，紧急情况下极易发生交通事故。
5、智慧家居
家中有非常多的智能家居的设备，智能家居不同产品之间互动场景的定义，需要边缘计算。另外，对于智慧家居来说，接入网络的安全性和私密性也为人们所看重，边缘云可以在物联网网关和数据中心之间建立加密通道，进一步提高系统的安全性和隐私性。
6、智慧城市
边缘计算就好比城市神经末梢，将人工智能与分布在城市中的传感器结合，可以高效处理城市运营问题，如在道路两侧路灯杆上安装传感器，收集城市路面信息，检测空气质量、光照强度、噪音水平等环境数据。
7、智慧路灯
嵌入到路灯内部的传感器、执行器、计算和存储单元可以组合起来构成边缘计算的节点，传感器采集的数据发送到位于网络边缘的计算和存储节点，经过计算将结果返回给执行器，执行器对路灯进行控制，而不是将数据发送到位于网络边缘的云计算中心。这样既可以提高系统的实时性，又可以减轻云端的压力。
8、风力发电
在风力发电机机组上布置边缘节点，实时收集数据信息。数据信息上传至工业网关，如风速、启动等做优化，将模型转化为算法或者规则，即时控制机组。
9、医疗保健
医疗设备上存储的数据可用于更新患者的数字医疗记录。边缘计算将连接起来这些医疗设备，在紧急情况下为医院和医生提供可靠和最新的患者信息。
10、无人机
边缘计算使无人机能够检查数据并实时响应数据，广泛应用于各种领域，如当无人机识别到车祸时，无人机可以向附近的行人提供有价值的信息。
六. 结语
据IDC预测，未来超过50%的数据需要在网络边缘侧分析、处理和储存。边缘计算将延伸至交通运输系统、智能驾驶、实时触觉控制、增强现实等诸多领域，成为运营商数字化转型的关键使能技术。