Edge computing （Edge Computing）—— The next explosive point of the Internet of things
Edge computing is 5G Business areas to be considered in the era of the Internet of things , It has a broad market prospect . As for the communication industry , Internet industry ,IT Friends interested in industry , Never miss the wave of edge computing . that
What is edge computing ?
<> What is edge computing ?
Edge computing (Edge computing
) It is a network edge side that is physically close to the data source , Converged network , calculation , storage , Open platform for application of core competence , Computing mode of providing edge intelligent service nearby .
The location where edge computation occurs is called edge node , It can be any node with computing resources and network resources between the data source and cloud center . such as , Mobile phone is the edge node between people and cloud center , Gateway is the edge node between smart home and cloud center .
In order to understand edge computing more vividly , Let's take a look at the inborn “ Edge computing ” Capable person —— octopus .
Octopus has a huge number of neurons , But there are 60% On the octopus's eight legs （ Brachiopod ） upper , The brain has only 40%.
In other words, octopus has a brain and multiple cerebellums similar to distributed computing . The brain is like a cloud center node , The cerebellum, the octopus' claw, is similar to the marginal node .
Edge computing is also a kind of distributed computing , Data processing , The running of application and even the realization of some function services , From the network center down to the node on the edge of the network , Processing data nearby , There is no need to upload a large amount of data to the remote core management platform .
<> Why edge computing is needed ?
The Internet of things is booming in various fields , The era of the interconnection of all things is approaching . But with the development of the business , Rapid growth of Internet of things devices , It is gradually found that cloud based computing can not meet the actual needs of many scenarios .
<>1. Massive data brings great pressure on network bandwidth
Cloud center has powerful processing performance , It can handle massive data . however , With the development of Internet of things , Almost all electronic devices can now be connected to the Internet , These electronic devices will produce huge amounts of data , It is a problem to transfer these massive data to the cloud center , This is a huge challenge for network bandwidth .
<>2. For low networking devices , Increased demand for collaborative work
The system performance bottleneck of cloud computing model lies in the limitation of network bandwidth , It takes time to transfer massive data , Cloud centers also need time to process data , This will increase the response time of the request , Poor user experience . And in the emerging Internet of things application scenarios , For example, real-time voice translation , Another example is driverless cars , There are very high requirements for response time , Relying on cloud computing is not realistic .
<>3. Networking devices involve personal privacy and security
A large amount of data in terminal devices will involve personal privacy , Data security risk will be greatly increased when it is transmitted to the cloud center .
If you can be like an octopus , The method of edge calculation is adopted , Massive data can be processed nearby , A large number of devices can also achieve efficient collaborative work , Many problems can be solved easily . therefore , Edge computing theory can satisfy many industries in agility , Real time , Data optimization , Applied Intelligence , And the key requirements of security and privacy protection .
<> Types of edge computing
When deploying edge business , There are three types of edge scenarios to consider . The difference is personal edge （personal edge）, Business edge （business edge）, Cloud edge （cloudy
<>1. Personal edge
Around the edge of our personal computing , Sometimes it's right next to us , It's in our house ; For example, smart phones , Home robot , Smart glasses , Medical sensors , Wearing a watch , Smart speaker , Other home automation systems .
Personal edge devices are generally mobile , So it is also called moving edge computing （MEC), Endurance needs to be considered , Characteristics of network handoff and offline conditions .
<>2. Business edge
Service edge is used to gather information of personal edge devices , robot , Information such as sensing equipment is collected and processed here . Such devices can be deployed in office or home areas , It is used to support the information concentration within the region , interactive , handle .
<>3. Cloud edge
A complex Internet of things application , It will involve the collaboration of multiple cloud platforms . speech processing , Face recognition , The rise of vertical cloud platforms such as medical artificial intelligence , It improves the intelligence of the Internet of things , However, there is a higher demand for cooperation between platforms . Cloud edge is equivalent to providing data analysis on different cloud platforms , Data interaction , The function of data collaboration .
<> The advantages of edge computing
Compared with the shortcomings of cloud computing in the development of the Internet of things , The advantages of edge computing are obvious . put in plain language , Edge computing is for the pain point .
<>1. Save heart, effort and flow
* A large number of computing tasks can be processed near the source of data generation , This greatly eases the transmission pressure of the network .
* It also reduces the pressure on the data center .
<>2. Real time, fast and efficient
* Edge computing has the advantage of real-time processing because it is distributed and close to the device , So it can better support the real-time processing and execution of local business .
* Tasks are processed at the edge , therefore , The time of data transmission in the network is sharply reduced , Users get a response faster .
<>3. Intelligent safety, more energy saving
In a family or other relatively private environment , We generally don't want our privacy data uploaded to the cloud center , Cloud center to solve related problems . We hope that this kind of computing task can be solved locally , This can better protect privacy data .
* After offloading some computing tasks from the cloud to the edge , The energy consumption of the whole system has been reduced 30%-40%.
* Using multiple edge nodes to cooperate , That is to ensure efficient problem solving , At the same time, it can balance the problem of data privacy and the cost of data transmission in the network
<> The challenge of edge computing
Everything is a double-edged sword , Edge computing has obvious advantages , Naturally, there will be no small challenge .
<>1. General computing capability of edge nodes
theoretically , Edge computing can be done on several nodes between the edge device and the cloud platform , Including access points , base station , gateway , Business node , Router , Switch, etc . However, due to the heterogeneous platform , The edge of every network is different , The computing task is divided into edge nodes of different platforms . The running time of different nodes is different , Developers are faced with great difficulties . How to develop portable solutions across different environments , To support common computing requirements is a major challenge .
<>2. Find the edge and assign tasks
reach 2020 There will be 500 100 million terminals and equipment networking , In addition to the edge devices and terminal networking largest “ isomerism ” Beyond the characteristics , Product life cycle is getting shorter and shorter , The demand of individuation is higher and higher , The trend of life cycle management and service is more and more obvious , These new trends need strong technical support from edge computing .
How to discover resources and services in distributed computing environment , And the effective allocation of tasks is an area to be expanded . In order to make full use of network edge devices , Some kind of discovery mechanism needs to be established （ Such as naming mechanism , Network protocol, etc ）, Find the right nodes for decentralized deployment , Dynamic implementation , Large scale deployment of computing and storage capabilities and efficient collaboration between cloud and device , Seamless connection .
* These mechanisms must not increase the waiting time or damage the user experience , Seamless integration of different levels and levels of computing workflow , The original mechanism based on cloud computing is no longer applicable in the field of edge computing .
<>3. Data storage and management
In the Internet of things environment, there will be a lot of data generation , The data should be read, written and manipulated by the application , Because of the heterogeneity of things in the Internet of things , The generated data is in various formats , Formatting all kinds of data is a challenge for edge computing .
The storage capacity of edge nodes is limited , So how do edge nodes handle this data , And how to save the data becomes the key to the problem . If you filter out too much raw data , It will lead to the unreliability of edge node data report , If you keep a lot of raw data , Then the storage of edge nodes will be a new problem .
<>4. Service quality (QoS) And service experience (QoE)
The service management of edge nodes needs to ensure an efficient and reliable system . Need to ensure high throughput of edge nodes , And it runs reliably when taking on additional calculation workload ; High risk nodes can be detected in advance , So as to avoid the risk that the loss of nodes may lead to the unavailability of services ; Nodes should be able to exchange status and diagnostic information , Ensure the reliability of data in sensing and communication .
Different services should have differentiated priorities . such as , Key services such as judgment and fault alarm should be higher than other general services , Services related to human health, such as heartbeat detection, have a higher priority than entertainment related services .
Things in the Internet of things are dynamic , It's not easy to add or delete an item to the Internet of things , The lack of service or whether a new node can adapt to it is a problem to be solved , These problems are of great importance to the edge OS High scalability and flexible design is a challenge .
<>5. Open and secure use of edge nodes
Security across cloud and edge computing , End to end protection is required . Because it's closer to the Internet of things , Therefore, the breadth and difficulty of network edge access control and threat protection have been greatly improved . Edge side security mainly includes equipment safety , network security , Data security and Application Security . in addition , Integrity of key data , Confidentiality is an important content in the field of security .
If the terminal equipment ( For example, switches , Routers and base stations ) As a shared access edge node , Many problems need to be solved ：
* The risks associated with the users and owners of edge devices need to be defined .
* When the device is used for edge computing nodes , The original function of the equipment cannot be damaged .
* Multiple users on edge nodes need to take security as the primary concern index .
* It is necessary to guarantee the minimum service level to the users of edge nodes , Set permissions for different applications , Restrict access to private data .
* Workload needs to be considered , Computing power , Data location and migration , Maintenance costs and energy consumption , In order to establish a suitable pricing model .
<> Opportunities and challenges coexist
The development of Internet of things and the bottleneck of cloud computing promote the rise of edge computing , Processing data at the edge node can improve the response speed , Reduce bandwidth , Ensure the privacy of user data . Edge computing is still in its infancy , It is possible to pave the way for more efficient distributed computing . Although there are many challenges in the implementation of edge computing , But edge computing will give birth to more development opportunities .
* standard , Benchmark and market ： Unified data connectivity and data aggregation are the foundation of business intelligence , Facing the diversified and heterogeneous technologies and standards existing in the current industrial field , Cross manufacturer is indispensable , Cross domain data integration and interoperability .
* Architecture and language ： With the increasing number of edge nodes supporting general computing , The need to develop frameworks and toolkits will grow .
* Lightweight libraries and algorithms ： Due to hardware limitations , Edge nodes do not support large software , Edge analysis needs lightweight algorithm , It can do reasonable machine learning or data 处理任务.