The ultimate goal of the Internet of Things is to realize the Internet of Everything in the future. We all know that the Internet of Things is composed of networks, and the network is composed of data. Therefore, big data plus cloud computing is an important foundation for the Internet of Things. Our next step is It is the development of the calculation of fog to the direction of Internet of Things computing.

At present, the development trend of the Internet of Things industry mainly focuses on three main lines: one is the industrial and commercial Internet of Things to B, the other is the government-driven field related to smart cities and outdoor infrastructure, and the third is the household Internet of Things. Relative to users, the three main lines are the most concerned about families and individuals, but the short-term trend is the fastest to the Internet of Things service.

The development trend of the Internet of Things can be seen from the trend of enterprise Internet of Things, and the integration of enterprise Internet of Things is phased. The first phase is the electronic product without Internet connection, and the second phase is the foundation of the Internet of Things, that is, connecting devices to the Internet, and remotely controlling and collecting data. This is what Wit Cloud has been doing for the past few years. The third stage is serviceable, through data collection to understand customer needs, while providing better services. The fourth stage is intelligent, which means that the device can judge by the algorithm, instead of simply using the mobile APP. The fifth stage is to achieve continuous optimization, because the Internet of Things itself is also an Internet product. After the integration of the Internet with a company, it has the opportunity to open up its upstream and downstream industries (including the upstream supply chain and downstream distribution channels). I am able to optimize my operation. The sixth stage is differentiation. With the development of various sub-sectors, the business model of many enterprises is constantly changing. Therefore, continuous optimization is an important opportunity that the Internet of Things can bring to enterprises. Basically, continuous optimization and the formation of differentiated business models are things. The long-term goal of networking development.

Fog calculation is to add a layer of "fog" between the terminal device and the cloud data center, that is, the network edge layer, such as adding a small server or router with memory, and putting some data that does not need to be put into the cloud. Layers are processed and stored directly. Fog calculations are more geographically distributed and have a wider range of mobility, which allows it to adapt to the growing number of smart devices that do not require a lot of computation. For some time-sensitive applications such as real-time and streaming applications, fog computing has a greater advantage. For example, the thermometer's readings per second are not required to be uploaded to the cloud. What fog computing technology needs to do is to get an average based on real-time data and then upload it to the cloud every half hour or so. If the temperature is abnormal, the Sensor can still respond quite intelligently and quickly.

Fog computing expands the network computing model of cloud computing, extending network computing from the network center to the edge of the network, and thus more widely used in various services. By using fog calculation, the calculation and storage pressure of the cloud can be greatly reduced, the efficiency is improved, the transmission rate is improved, and the delay is reduced.

In the early years, Cisco proposed "fog computing", which is an extension of "cloud computing", which is used to promote its product and network development strategy. It hopes that it will not be constrained by cloud computing and how to store and process it on IoT devices. The data they produce themselves. However, since Cisco proposed fog computing to the present, there are not many cases of real landing. Cisco's router and switch technology lacks the definition and control capabilities of terminal devices. Most of the scenarios cannot truly reflect the ability of fog computing.

Over the past 20 years, the rise of cloud computing has moved the end of computing and storage to the cloud, facilitating data aggregation and unified management. Now, "fog computing" distributes the calculation and storage of aggregations to pipes and ends. Faster response and ultra-large-scale computing system. This new computing system puts forward new requirements for a company's control over cloud, management, and dynamic management capabilities.

The calculation of fog in the commercial use of the Internet of Things has great advantages, mainly in four aspects.

1. Reduce energy consumption. Cloud computing puts a large amount of data into the "cloud" for calculation or storage. The core of the "cloud" is a "data center" with a large number of servers and storage. Since semiconductor chips and other supporting hardware are still very power-consuming, the power consumption of the global data center is equivalent to the power supply of 30 nuclear power plants, and 90% of the power consumption is wasted because the efficiency is very low. Google’s global data center has reached 300 million watts of electricity, which exceeds the electricity consumption of 30,000 American households. When the amount of data transmission in the future increases further exponentially, the cloud center will no longer be able to sustain it. This data transmission refers to the transmission between a large number of wireless terminals and the "cloud".

2, improve efficiency, with the arrival of the Internet of Things, all walks of life including home appliances, wearables, automotive peripherals, industrial agriculture, commercial equipment, and other types of terminal equipment that need to be connected to the network, will generate extremely large amounts of data Sending and receiving may cause I/O (input and output) bottlenecks between the data center and the terminal. The transmission rate is greatly reduced, and even a large delay is caused. Some devices that need real-time response cannot guarantee normal operation, for example: None Man-machine, security alarm, monitoring equipment, etc.

3, data analysis, a large number of enterprises for the massive data collection needs of the solution is to reduce the frequency and total amount of data collection, such as sampling every 10 minutes, may be collected hundreds of times a day, accuracy and efficiency will be greatly reduced, some Devices that require massive, uninterrupted data collection will reduce their service value, and some devices that need timely decision-making will wait for all data to be uploaded to the cloud computing decision before returning to the device side, which will greatly reduce service capabilities.

4, upgrade security, in the absence of a mature technology platform, how to calculate most of the equipment, has been finalized at the factory, unless a very heavy way to remotely upgrade its entire system, but this upgrade is inefficient and dangerous It is possible to change the operating system, and millions of devices on the market will never lose contact.

With more than 6 years of experience in IoT technology accumulation and innovation, Wisdom Cloud supports more than 7 million IoT devices. It has been at the international leading level in technology innovation and application. With the continuous development of IoT industry, networked devices Computing power is also increasing, and wit cloud pushes the application of “fog computing” further in the IoT field to meet the specific functional requirements of customer products. As a full stack IoT Platform, Wit Cloud has strong control over cloud, management and end. This is a very special advantage for platform landing fog calculation.

The wit cloud's landing strategy for fog computing is different from Cisco and other communications equipment vendors. Cisco mainly relies on routers and switches to deploy fog computing, but these "pipelined" devices are suitable for data handling, data cannot be parsed, and can not affect the operation of smart terminals and the cloud. It is difficult to apply "fog computing". . The power of the "fog calculation" of the wit cloud now falls on the larger number of communication modules DTU (Data Transfer Unit) and Gateway (Networking Gateway). These devices are all directly controllable computing units of the wit cloud. "Fog computing" is the dynamic allocation of computing and storage capabilities from the cloud to these edge devices.

Another innovation of the wit cloud is to give the ability to dynamically deploy computing to low-level processors, and even a few dollars of ordinary microcontrollers (MCUs) can participate in fog calculations. The wit cloud ECE fog computing framework embeds a "micro-container" on the DTU or gateway that can execute "micro-applications" composed of lightweight scripting languages ​​such as Javascript, Python, and Lua. These "micro-applications" can do data processing, protocol conversion, and interoperability between devices. Developers can write various scripts directly in the cloud, which can be easily pushed to the micro-application container through the ECE system, and the computing power can be deployed to the device in real time. The device does not need to be restarted, and the original system OTA firmware upgrade is not required. Instead, you only need to update its algorithms and micro-applications on the fog side.

Devices with “micro-applications” can realize millisecond-level data acquisition and analysis according to business needs, provide greater analyzable data volume, and have local judgment capabilities, and the accuracy and efficiency are greatly improved. However, the computing on the device side does not replace the cloud computing. Instead, the device end processes the micro-application data, sends the processed useful data to the cloud, and the cloud then performs data aggregation. The device end effectively amplifies the cloud aggregation capability. . Now, through the way of fog calculation, the cloud can quickly and flexibly update its micro-application on the fog side, effectively upgrading the "software-defined hardware" to "cloud-defined hardware."

The Internet of Things generates a lot of data every moment, and when the data is more and more, it needs to be processed by the machine. This has to mention artificial intelligence, deep learning, existing human-computer interaction companies, and people like Apple Siri. Less, so artificial intelligence can really achieve business value. The future is able to create value for the enterprise is a good technology, so the future of artificial intelligence certainly can not just catch the eye.

The algorithm in artificial intelligence is understood from the time dimension or the inclusion relationship, and can be divided into three layers: one is pan artificial intelligence. A variety of algorithms for simulating human behavior to identify, make decisions, plan, and predict can be collectively referred to as artificial intelligence; second, machine learning. It is a branch of artificial intelligence. It is a process of storing and analyzing a data and multi-dimensional data for a long time, allowing the machine to self-integrate and mine the inherent regularity of the data. This is the field of machine learning; the third is deep learning. In fact, it is aimed at the further evolution of neural networks in machine learning. Emphasis is placed on the construction of a complex and deeper neural network model by simulating the human brain. Different from simple shallow learning, the network of deep learning can reach several hundred layers or even thousands of layers. The resulting model's ability to understand and understand the data is far beyond machine learning.

For the company that makes the tool, it is to turn the complicated principle into a functional module, so it defines the field of the Internet of Things. In the application scenario just mentioned, which algorithms are more general and practical, including scheduling and anomaly detection. There are also recommendations, dimensional reductions, which are more complicated. When you have a lot of data, in order to get decisions faster, you need to reduce its dimensions. This is the methodology of artificial intelligence, you can turn it into some Standardized module. These algorithms can be used to toolize, define artificial intelligence as the direction of the tool, turn the algorithms used in the industry into one tool module, and then combine.

In the early years, Google acquired Nest, the core technology advantage of Nest is to quickly transform the traditional thermostat, and all the air conditioning equipment in the home can be combined with the heating and cooling equipment. By knowing the characteristics of the energy used in the home, it is possible to predict when to put the home. To what extent the temperature is controlled, this method can save energy at home.

How to apply? There are three more important problems in water heaters. The first is that electric water heaters are slower in hot water. The second is that the heating energy consumption of different manufacturers is not the same. The processor in the water heater is not like a dual-core or quad-core inside the drone. It is just a simple MCU and a simple processor. Now the ability to give it communication, can be connected to the wit cloud, combined with algorithms to make it more energy efficient and more convenient to use.

How to save energy? First, we need to make a user model, know when to use how much water, the general household water heater does not have a water sensor, do not know how much water it uses, can only use the algorithm to predict it according to temperature, according to complex models, multiple dimensions After the data is aggregated, it is concluded that the shape of the water used by the family in the past month looks like.

The next step is to start making predictions. The goal is to predict the likelihood of water use and water usage for any minute tomorrow through a month of study. That is to say, the detection of water is simulated by an algorithm. Through the construction of the user model, the recommendation engine and the prediction engine are used to make decisions, so that the product can automatically run the optimized algorithm.

In the future, the terminal devices of the tens of billions of Internet of Things are so-called lower-level creatures. The sensors inside may be low-power, and the computing unit is also low-power. It is very troublesome to do complex deep learning. The current practice is to use data collection with lower organisms, pre-processing part of the budget on lower organisms, allowing pre-processing and feature processing to be handled here, reacting faster in the cloud, and future tens of billions of devices will It is getting smarter and will not be affected by the instability of the network.

Machine learning on the wit cloud platform is called Giga Machine Learning, which can be distributed to millions, tens of millions, or even hundreds of millions of billions of devices. Put some algorithms into the SKU of low-lying creatures, the low-frequency collection frequency is very high, if you do pre-processing, you can throw the feature values ​​into the cloud to do the aggregation process, the algorithm updates in real time, push the algorithm to the end, the data is on The cloud is dynamically combined with other data.

Modern Physics Experiment Series

Modern physics experiment related equipment for efficient specialized physics laboratory

Modern Physics Experiment Instruments,Optical Instruments,Acousto-Optic Modulator Experimental Device,Optical Spectroscopy Experiment Determinator

Yuheng Optics Co., Ltd.(Changchun) , https://www.yuhengcoder.com