Lei Feng network (search "Lei Feng network" public number attention) by: The author of this article Xia Wei, a French telecommunications doctoral student, pays attention to fog computing, cloud computing, distributed computing.
With the advancement of science and technology, various technical terms are also rapidly developing. Cloud computing, Internet of Things, and fog computing, I believe these terms have long been in everyone's view. Among them, the concept of fog computing was proposed by Cisco in 2011 [1] and it is relatively new. Today, fog calculation has become a hot topic and focus of research, and has been high hopes for the industry.
However, the author still can see that there are no doubts about the fog calculation. There are two things:
Is not a local server?
In the fog, hype it?
The calculation of fog is far more than simple. It is a consideration of the integrity of tens of thousands of “local serversâ€. It is a platform, not a single machine. In the follow-up content, we will discuss in depth.
Take a look at the direction chosen by the world's tech giants. All of the following items have a great correlation with fog computing. They spend large sums of money on hiring top scientists to hype?
OpenFog [2], a fog computing research project co-funded by Arm, Cisco, Dell, Intel, Microsoft, and Princeton University.
Orange (France Telecom) and the Inria (French National Institute of Computer and Automation) led the joint research of fog computing and large-scale distributed cloud research [3].
Huawei's "full cloudization" strategy [4]
Intel's "Cloud Computing at the Edge" project [5].
NTT's "Edge Computing" project [6].
AT&T's "Cloud 2.0" project.
For this question, I hope to use this paper to reveal the significant value of fog calculation.
Speaking from the Internet of ThingsSmart cities, smart homes, and various predictable applications for the Internet of Things will greatly facilitate people's lives in the future. However, the intelligence of smart terminal devices on the market is generally unsatisfactory. So where does this "smart" come from? How can we guarantee the intelligence of the equipment?
The basis of computer intelligence lies in the resources behind it, such as CPU, memory, hard disk, network bandwidth and other computing resources (more precisely, the CPU and memory are categorized as computing resources; hard disks are attributed to storage resources; bandwidth is attributed to communication resources. For the sake of simplicity, this article refers to them uniformly as computing resources; video, temperature, light intensity, and other data sources provided by sensors; and of course, power.
The core of these resources is computing resources, which extract knowledge and make decisions through calculations; save knowledge bases through storage to ensure accurate decisions and make predictions based on historical experience; and complete communication between devices through communication. Realize the distribution of knowledge and decisions. Based on the above, we can give users intelligent services and experience.
So now the device is not smart enough, the crux of whether the device's CPU is not strong, the memory hard drive is not big enough?
The lack of terminalWe cannot imagine installing a base station on every mobile phone. Similarly, we cannot imagine that every device has a lot of resources. This will significantly increase the cost and it will not be able to form an effective solution.
When resources are scarce, an intuitive idea is to hand off computing tasks to other computing-powered devices. There are a large number of terminal devices in the Internet of Things. They can't make decisions in the local computing. Who should solve the problem of insufficient resources for the terminal devices? Everyone thought of the cloud.
The lack of cloudThe cloud computing platform provides cloud users with resources in the data center. In the past ten years, cloud computing has fully demonstrated its superiority to people:
"Infinite" resource pool
A large number of users share resource pools for cheap resources
Access with any network device anytime, anywhere
"Fast" redeployment, flexible resource hiring
On-demand purchase, self-service
The service provider deploys a specific service in the cloud, and the terminal device sends the information to the service. After the service completes the calculation, the result is sent back to the terminal, and the necessary data is stored in the cloud. Through this form, the cloud fully meets the resource expectations of the terminal equipment and becomes an indispensable part of the Internet of Things ecosystem.
In order to serve users in different geographical locations, in a multi-level structure of the Internet, the data center is located in the core network. The core network is far away from the end user, and the user message needs to go through several hops before it can reach it. The following figure is a simplified part of the network topology.
Figure 1 Internet network topology
The data center provides a large amount of highly concentrated resources, but only the cloud still has some deficiencies.
High latency: Distances far from users result in higher network delays. Applications with high real-time requirements are difficult to deploy in the cloud.
Network Congestion: According to Cisco's forecast, by 2020, there will be 50 billion smart devices worldwide. In contrast, the growth rate of network bandwidth is far behind. If a large number of IoT applications are deployed in the cloud, there will be a large number of sensor raw data constantly flooding into the core network, causing core network congestion.
Lower reliability: security, life-related applications for Internet of Things, once the application fails, the data center fails, or any segment of the network from the end user to the cloud platform fails, it will bring major security risks. The communication path from the terminal to the cloud is long and the risk of failure is high. The cost of deploying service backup in the cloud is also high.
It can be seen that the cloud is not suitable for applications that require real-time, big data, and high reliability. People need new computing models to meet future applications and make up for cloud deficiencies. It is in this context that fog calculations are presented.
Fog calculationThe fog calculation is a very visual name, and Ginny Nichols proposed it with an interesting statement: "The fog is a cloud close to the ground."
This sentence has two meanings:
There are many similarities between cloud computing and cloud computing. For example, they are based on virtualization technology and provide resources for multiple users from a pool of shared resources.
Close to the ground. This also points to the first difference between fog and cloud - location. More specifically, their position in the network topology.
Figure 2 Original definition of fog calculation
The figure above is based on Cisco's original definition of fog calculation [1]. In Cisco's definition, fog mainly uses devices in the edge network. These devices can be traditional network devices (routers, switches, gateways, etc. that have long been deployed in the network), or they can be specially deployed local servers.
In general, devices that are specifically deployed will have more resources, while traditional network devices that use ample resources can significantly reduce costs. The resource capabilities of these two devices are much smaller than a data center, but their large number can compensate for the lack of a single device resource.
The fog platform consists of a large number of fog nodes (ie, the hardware devices used in the above fog and the management systems within the devices). These fog nodes can be scattered in different geographical locations, which is in stark contrast to the centralized data center.
Based on the above, the difference between fog computing and cloud computing can be summarized:
Lower: The fog node has a lower position in the network topology and has a smaller network delay (total delay = network delay + computational delay) and is more responsive.
More: Compared to the cloud platform's constituent unit, the data center, there are a large number of fog nodes.
More extensive: fog nodes have a wide geographical distribution.
Lighter: fog nodes are lighter and computing resources are limited.
What are the advantages of these different kinds of fog and what makes it an integral part of the Internet of Things ecosystem?
The advantages of fogIn addition to the low latency mentioned above, fog calculations have the following advantages:
Provincial core network bandwidth: Fog acts as a middle layer between the cloud and the terminal, and it is in the communication path between the user and the data center. The fog can be filtered, aggregate user messages (such as sensor messages that are sent continuously), and only send necessary messages to the cloud to reduce the pressure on the core network.
High reliability: In order to serve users in different areas, the same service will be deployed on the fog nodes in each area. This also makes high reliability an intrinsic attribute of fog calculations. Once an area's service is abnormal, user requests can quickly switch to other nearby areas.
Background information: Because in different regions, the service in the fog calculation can understand the regional background information, such as whether the bandwidth in the region is tight. Based on this knowledge, a video service can timely decide whether to reduce the video quality of the region to avoid Caton arrival; and for a map application, you can cache the map of the area to improve the user experience.
Power Saving: The power consumption of data centers has become an important cost, and the cooling system occupies a significant proportion. Because of the geographical spread, the fog computing nodes do not generate large amounts of heat and do not need additional cooling systems, thereby reducing power consumption.
Based on the above advantages, fog can make up for the lack of cloud and cooperate with the cloud to work together.
Cloud + fogFog computing is proposed as an extension of cloud computing rather than an alternative to cloud computing. As mentioned above, in the IoT ecosystem, fog can be filtered and user messages can be aggregated; user data can be handled anonymously to ensure confidentiality; data can be processed initially to make real-time decisions; temporary storage can be provided to enhance user experience.
In contrast, the cloud can be responsible for large amount of computation, or long-term storage tasks (such as: historical data preservation, data mining, state prediction, overall decision-making, etc.), so as to make up for the shortage of single fog nodes in computing resources.
In this way, cloud and fog together form a computing model that benefits each other. This new computing model can better adapt to the IoT application scenario.
ExampleThe current urban road monitoring system, from the monitoring probe to the local center computer room communication hops is generally 3 to 4 hops or even higher, if the system needs to make real-time decisions will face the challenges of network delays.
Figure 3 Use Case - Intelligent Traffic Light System
The figure shows an intelligent traffic light system with the exception of the monitoring probe as a sensor and the traffic light as an actuator. The introduction of fog calculations will bring more possibilities to this system. Such as:
In the monitoring process, compared to the previous frame, usually only part of the picture changes, and the other part remains unchanged, which is very suitable for compression processing. For images that require human monitoring, fog nodes directly forward the video stream to the central computer room; while other surveillance videos only need to be stored, they do not require high real-time performance. They can cache several frame images at the fog node, and transmit the compressed data to the central computer room. . This will ease the network bandwidth from the fog node to the computer room.
At the fog node, it can be determined whether there is an ambulance headlight flickering in the monitoring picture, and a real-time decision is made to send to the corresponding traffic light to assist the ambulance to pass.
The above example is just a concrete microcosm in a smart city. The application scenarios of fog computing in the smart grid, car networking, smart homes and other fields are numerous.
challengeWhile fog computing brings new possibilities, it also brings new challenges in terms of security, efficient use of resources, and APIs. Fog uses a large number of decentralized devices to make centralized control difficult. Fog node resources are relatively limited, and coordination between nodes is required to optimize the deployment of each service. “When to migrate services to where Mobile terminal equipment, dynamic application scenarios need to consider the issue.
With the development of fog computing concepts, fog has been further extended to the "ground." The fog node is no longer limited to the network edge layer, but also includes terminal equipment with ample resources.
Figure 4 fog development development definition icon
The terminal device interacts with the user directly and has a large number. While enriching the types of fog devices, it also brings more dynamic attributes such as battery power and fog node mobility to be solved.
ConclusionThis article starts from the application scenario of the Internet of Things, talking about the demand for cloud by the resource limitation of the terminal device, and the fog caused by the position of the cloud in the network. Together with everyone, we discussed the comparison of cloud and fog, the combination of clouds, the advantages of fog, the application of fog, and the challenge of fog. I hope that this article will be launched in order to attract the attention of everyone and pay attention to the development trend of science and technology.
references
[1] Fog Computing and Its Role in the Internet of Things. Flavio Bononi and al, Cisco. ACM SIGCOMM International Conference on Mobile Cloud Computing, August 2012.
[2] http://
[3] http://beyondtheclouds.github.io/dcc.html
[4] Huawei promotes "full cloudization" strategy, enabling digital transformation of the industry
[5] Increasing Network ROI with Cloud Computing at the Edge. Intel Solution Brief, 2014.
[6] Announcing the "Edge computing"concept and the "Edge accelerated Web platform"prototype to improve response time of cloud. NTT Press Release, January 2014.
Copyright Notice: CC Licensing Agreement: Keeping Signatures - Forbidden Deductions, Original Address: "Undoubtedly Cloud Computing" .