Current Status and Pain Points of Cloud Computing Industry

1.1 Cloud Computing Concept

Cloud computing is the growth, usage, and delivery model of Internet-based services, usually involving the provision of dynamically scalable and often virtualized resources over the Internet. The widely accepted concept of cloud computing at this stage is defined by the National Institute of Standards and Technology (NIST): cloud computing is a pay-per-use model that provides available, convenient, and on-demand network access, Access to a shared pool of configurable computing resources that can be rapidly provisioned with minimal administrative effort or interaction with service providers.

Cloud computing is the product of the integration of traditional computer and network technologies such as distributed computing, parallel computing, utility computing, network storage, virtualization, and load balancing.

There are many ways to classify the cloud computing industry. According to the type of service provided, it can be divided into three categories: IaaS, PaaS, and SaaS. From IaaS to SaaS, it is getting closer to "fool" software, which is convenient for users to use directly. Therefore, if the improvement of hardware usage efficiency and cost reduction of technological innovation is more reflected in the IaaS level, SaaS expands the market through price reduction (annual fee lowers the threshold of use) on the basis of enjoying hardware improvement.

1. The full name of IaaS is "Infrastructure-as-a-service" (Infrastructure-as-a-service), which provides underlying infrastructure resources such as servers, storage, and network hardware. After purchasing IaaS products, users must complete the environment configuration and application development by themselves. It is difficult for commercial customers to use directly, and most of the users are software developers, especially PaaS and IaaS product developers;

2. The full name of PaaS is "Platform-as-a-service" (Platform-as-a-service), which provides software deployment platforms, such as virtual servers and operating systems. Users do not need to pay attention to the bottom layer, but only need to develop applications according to their own logic. Suitable for large commercial customers with clear characteristics and high IT budget, or application developers;

3. The full name of SaaS is "Software-as-a-service" (Software-as-a-service). It provides software that can be used directly.

IaaS, PaaS, and SaaS providers can cross boundaries with each other. At present, IaaS vendors can generally carry out further resource packaging, provide database, application middle layer package runtime, etc., to form a public PaaS platform, such as Amazon AWS. Manufacturers that provide SaaS, while providing common SaaS products for general commercial customers, will also create private PaaS products with their own characteristics for some large commercial customers, and even have some of their own IaaS products, such as Oracle.

1.2 Analysis of Three Models of Cloud Computing

first level title

The relationship between the three categories

Large enterprises mostly adopt private cloud, public cloud, and hybrid cloud deployment methods, while small enterprises mostly adopt public cloud, private cloud, and community cloud; since PaaS is in the middle of the cloud computing industry, IaaS providers can rely on hardware and technical advantages to provide For PaaS services, SaaS providers can also rely on customer relationships to provide PaaS services for large customers, and pure PaaS providers are not very competitive. Therefore, in the cloud computing industry, IaaS and SaaS have greater opportunities.

China's IaaS market is developing at a high speed. In China's IaaS market, Alibaba's market share is about 50%, and its cloud computing revenue growth rate can represent the growth rate of China's IaaS market. Alibaba's cloud computing revenue growth rates in 2015, 2016, and 2017 were 64%, 138%, and 121%, respectively, indicating to a certain extent that China's overall IaaS market is on a high-speed development channel.

R&D expenses are the biggest threshold in the IaaS industry, and most of them are controlled by giants

High R&D costs and large hardware costs can only be paid by large companies, whether domestic or foreign, IaaS is controlled by giants; and because IaaS development is strongly dependent on technology updates, high-income companies have capital investment in high R&D Fees, so the Matthew effect in this industry is very obvious.

According to the data, the top six IaaS public cloud market share in 2016 are Amazon, Microsoft, Alibaba, Google, Rackspace, and IBM. Except for Rackspace, which entered the IaaS industry in 1998 and has a large scale for a long time, other companies are all It is a giant in other industries and can pay huge R & D investment. However, Rackspace's 2016 IaaS public cloud revenue increased by 5%, and the revenue growth rate was much lower than that of the other five companies.

Amazon is the world's largest cloud computing company. The cloud computing AWS was launched in 2006. It mainly provides IaaS products and also provides some PaaS products. Most of Amazon's research and development expenses are used for cloud computing. By 2017, Amazon's total research and development expenses have reached 22.6 billion US dollars, and the year-on-year growth rate is as high as 41%.

Alibaba is the largest cloud computing company in China. In July 2015, it announced a strategic investment of 6 billion yuan in Alibaba Cloud for international business expansion, cloud computing, and big data field foundation and cutting-edge technology development; According to the report, it is expected to invest 100 billion yuan in the next three years to establish a research and development center "Dharma Institute", including the research and development of cloud computing basic technologies.

1.3 Cloud Computing Market Size

The global cloud computing market continues to grow driven by technology and price. According to Gartner's data, the cloud computing market including IaaS, PaaS, SaaS, process services, and advertising marketing was 219.6 billion US dollars in 2016, and the overall scale is expected to reach 411.4 billion US dollars by 2020, with a compound growth rate from 2016 to 2020 The rate is 17%.

image description

China's cloud computing market is developing faster. Compared with the global market, the Chinese market started late, the market is small, and technology is accelerating to catch up with the world's cutting-edge technology. Coupled with the increase in the number of customers brought about by manual substitution, the growth rate of China's cloud computing market is higher than that of the global market. According to the data, the overall market size of Chinese enterprise cloud services (including IaaS, PaaS and SaaS) was about 51.5 billion yuan in 2016, and the market size will be about 136.6 billion yuan by 2020.

image description

image description

Data source: Gartner

image description

1.5 Industry Status and Pain Points

1. Cloud Computing Monopoly

The existing cloud computing market is extremely centralized, with market shares Google, Amazon (AWS), Microsot Azure, Alibaba Cloud and Tencent Cloud have monopolized the entire cloud computing market by relying on their highly centralized server resources. High profits, which in turn lead to high prices for computing power services

2. No incentives

BONIC (Berkeley Open Infrastructure for Network Computing) is currently the most mainstream distributed computing platform, which is used by many projects in mathematical physics and other disciplines, but it is formed based on the computing resources of volunteers distributed all over the world A distributed computing platform lacks enough volunteers to contribute computing power.

3. Insufficient computing resources

Although we see the prosperity of DApps in the future, the current general blockchain has very limited computing power to run DApps, and the existing cloud computing infrastructure cannot meet the needs of DApps, which require a completely decentralized infrastructure to run; storage capacity Insufficient and high read latency of the protocol, these require additional computing resources to meet more demanding applications.

4. The cost is too high

first level title

cloud computing related concepts

When it comes to cloud computing, we have to mention two concepts related to cloud computing: fog computing and edge computing.

2.1 Edge Computing

Edge computing (Edge Computing) has long been proposed to represent the boundary of clouds and devices. It can be traced back to 2003, when AKAMAI and IBM began to cooperate to provide edge-based services on WebSphere servers.

Edge computing is a distributed open platform that integrates network, computing, storage, and application core capabilities on the edge of the network close to the source of objects or data. Key requirements for optimization, application intelligence, security, and privacy protection.

Edge computing will be a new ecological model. By converging network, computing, storage, application, and intelligence resources at the edge of the network, it can improve network service performance and open network control capabilities, thus stimulating a new model similar to the mobile Internet. business. The technical concept of edge computing has nothing to do with specific network access methods, and can be applied to different scenarios such as fixed Internet, mobile communication network, consumer Internet of Things, and industrial Internet to form their own network architecture enhancements.

2.2 Fog computing

Fog computing is a recent concept, pioneered by Cisco. Because compared with the cloud, it is closer to the place where the data is generated, and the data, data-related processing and applications are concentrated in the devices at the edge of the network, rather than stored almost entirely in the cloud. It expands the concept of cloud computing and is advocated by Cisco and others as a structure to realize IoT, aiming to be adopted globally.

Fog computing is a distributed computing model that serves as an intermediate layer between cloud data centers and Internet of Things (IoT) devices/sensors. closer to the sensor. The introduction of the concept of fog computing is also to deal with the challenges faced by traditional cloud computing in the application of the Internet of Things.

Fog computing and edge computing are vaguely defined, and the industry has been trying to separate the two as separate concepts. In this regard, the most widely accepted concept in the industry is that in edge computing, data processing is on the hardware that collects the data. Fog computing is when a subset of nodes send their data to a larger central connection point, where the data is processed while connected to the larger overall central network.

first level title

Blockchain Technology and Cloud Computing

3.1 The relationship between blockchain technology and cloud computing

Due to the popularization of the Internet, a large amount of multimedia information (graphics, text, audio, video, etc.) has created massive data, most of which are unstructured.

Distributed computing and parallel computing: Distributed computing is a solution that uses multiple hosts (clusters) geographically different to cooperate to complete a large number of computing tasks, thereby replacing supercomputers. Parallel computing refers to multi-CPU parallel processing. Parallel computing can improve computing efficiency, but the premise is that the program algorithm should be designed in parallel as much as possible.

Virtualization: Virtualization is a method of dividing resources for cloud computing, including two levels: physical resource pooling and resource pool management. There are two types of virtualization, one is to virtualize multiple physical resources into one "big" logical resource, and the other is to divide multiple "small" logical resources on one physical resource.

At present, cloud computing adopts the first virtualization method more often. The deployment method uses distributed clusters. Parallel computing does not involve much (parallel computing is still under research academically). The terminal group provides IT services on demand. It can be simply understood that the distributed solution is proposed to process massive amounts of data quickly, and the ultimate purpose or manifestation of the solution is to realize "exchanging hardware for time" by distributing data on multiple computers , to process simultaneously (parallel processing) datasets with certain attributes. From a reality point of view, many micro and small Internet companies do not have the ability or need to build their own distributed systems, and will rely on cloud service providers outside of local resources, so that companies can focus on product and business implementation.

Combined with flexible development tools, the DCC platform can help developers release software and make money, thereby changing the way computing tasks are organized and executed. Such platforms become the foundation for building the future Internet by enabling decentralized microservices and asynchronous task execution. Complex applications (such as CGI rendering, scientific computing, machine learning, etc.) will benefit everyone thanks to the dramatic reduction in the price of computing.

Computers are connected through a P2P network, allowing application owners and individual users (computing power "requesters") to rent computing power from other users (computing power "providers"). These computing resources can complete computing tasks that require certain computing time and computing power. At present, computing resources are controlled by centralized cloud service providers, subject to closed networks, external payment systems and rigid operating models. Decentralized cloud computing facilities can implement a payment transfer system based on Ethereum (or similar public chains), which can realize direct payments between computing power buyers (requesters), sellers (suppliers) and software developers .

3.2 Advantages of DCC

1. In addition to free servers, bandwidth and other resources, computing needs are distributed to many nodes in the system, and users' prophetic resources are utilized to create value.

2. By integrating idle computing resources in the society, it provides decentralized computing power services for enterprises or individuals, and builds a computer computing power buyer's and seller's market based on the Token economy.

3. Compared with traditional cloud computing services, it reduces the threshold and usage fees of cloud computing services, which is conducive to the popularization of cloud computing.

3.3 Business model

3.3.1 Token Economic Model

Although there are many distributed computing power platform projects on the market, the overall technical framework is similar. The following is a general framework to roughly describe the Token economic model.

request node

This node is generally served by some merchants or scientific research institutes that have computing needs. Generally, the number of computers in laboratories or self-owned computers does not meet their current computing needs. Supercomputing or other cost-effective Computing resources such as globally distributed computing power to achieve the goal. The requesting node may be required to first model its own requirement file (data) according to the specification before entering the network (ELastic), or have other nodes do this simple classification work.

The following nodes need to contribute computing resources

classification node

This node is participated by users who contribute CPU computing power. Through the distributed computing power platform, some relatively simple classification algorithms are set up to classify certain computing purpose projects to achieve the effect of data classification and model. The purpose of this is It is able to better process these originally irregular data and at the same time transmit it to processing nodes that specialize in processing this type of data.

processing node

This node is dedicated to the processing of specific categories of data. Since there are thousands of types of data from classification nodes or request nodes, the way specific data should be processed or the model that should be used is also different. For some more complex projects, for those who can participate in this project The machines and people in it have high requirements. Here are two examples:

In the field of medicine, such as some medical data such as image data, the processing of such data generally requires personnel with at least certain medical knowledge to process and calculate this part of the data to obtain the required data.

Validator

Validator

This node judges and screens the processing results of the same data transmitted from multiple processing nodes, and generally only requires the contribution of the device CPU. Generally, the same data processing task will be sent to multiple data processing nodes, so that a vote can be made later to judge which data meets the requirements. Although doing so will cause data redundancy, it will provide a high guarantee for the correctness of the final result, and cooperate with the honor mechanism to effectively solve the problem of data fraud.

3.3.2 Honor System

Since this platform will no longer distribute computing power through voluntary distribution, how to measure the computing power contribution of each person is a new problem. Points system and member ranks can be a better way to measure how much a user contributes.

There is a possibility to perform points according to the number of completed task units, but because the projects that can be run on the distributed computing power platform in the future may have great differences in purpose or operation, such as when a certain data packet is in a certain One machine takes about an hour to complete, while another data package takes 20 times longer to complete on the same machine, which will cause the same task unit to get the same reward but different workload Obviously, it is not feasible to use the number of completed tasks to measure the amount of computing contributions made by users. Similarly, it is not feasible to measure the contribution by the cpu time required to complete the processing of a task. Since it is not feasible to summarize the contribution of a user through some specific parameters, some algorithms need to be used to accurately record the amount of calculations actually completed by each user in order to achieve a fair and accurate distribution of rewards. Therefore, the consideration of the amount of contribution should be referenced from multiple angles.

3.3.3 Points system

A more feasible point system should comprehensively consider several aspects to calculate the contribution of a node:

device performance index

Devices with different performances have different resource usage and costs when running and processing the same data packet. A standardized performance test should be conducted on different devices to obtain a reasonable performance score for weighting.

The correct number of results submitted

In a distributed computing power platform, the demand side is most concerned about the quality of the results obtained. If the results obtained by the device do not meet the requirements or are invalid results generated by malicious users, the reputation loss to the platform will be huge. Then what can be done is to grade according to the number of correct results submitted, and the correct results should be rewarded and upgraded, while the wrong results should be downgraded based on punishment.

Possible problems:

Performance testing is not allowed

Especially when there are cross-operating system platforms, such as installing the Windows version client and the Linux version client on the same computer, the benchmark test results may be quite different.

easy to cheat

first level title

DCC project analysis

Blockchain-related cloud computing projects mainly use the distributed technology of blockchain to connect multiple scattered computer nodes to provide distributed computing resource leasing services.

4.1 List of DCC projects

4.2 Item-by-item comparison of key projects

first level title

DCC project investment logic

1. Innovation of project technology

Blockchain projects related to distributed computing have the same general idea. They provide decentralized computing power services for enterprises and individuals by integrating and utilizing idle resources, and build a computer computing power buyer's and seller's market based on the Token economy. Therefore, technological innovation is particularly important. It is easier to improve the technology as soon as possible, and realize the landing through the test network to highlight the advantages in the competition.

2. Efficient organization and coordination platform

Projects that coordinate computing nodes through the blockchain want to successfully correspond to platforms in subdivided fields. In addition to finding their own scenarios, they also need to compete with traditional cloud computing projects for efficiency. Choosing a more efficient and low-cost basic chain is an important part of its stand out. .

3. Innovation in economic incentive model design

The design of a good economic model determines the long-term operation of a project. Although there are many DCC projects on the market, their economic model frameworks are basically the same, and the homogeneity is relatively serious. Then, under the premise of ensuring the safety of the existing economic model, Only by being innovative in the incentive model and more in line with users and users can it survive the competition for a long time.

4. Community operation ability

For blockchain projects, the community is a very important resource. The strength of the team's community operation ability determines whether a network effect will be formed, thereby determining the promotion of the project and the improvement of distributed computing power; and the community can also Make considerable contributions to the development of the project, including operation and maintenance, technical support and other aspects.

5. Whether the service quality can reach the commercial level

first level title

Future Trends of DCC Projects

1. With the improvement of technology, the actual number of projects implemented has increased

Blockchain-based cloud computing projects are not mature and perfect in technology, and some of them are still running on the test network. The stability and speed are much worse than traditional cloud computing services. At the same time, these projects do not have typical application cases or service scenarios. , It is still far from large-scale commercial use. It is foreseeable that with the simultaneous development of blockchain technology and distributed computing related technologies, landing projects will appear rapidly.

2. The improvement of security is an important development direction

In the cloud computing model, user data is uploaded to the data center, and in this process, data security becomes an important issue. From electronic financial account passwords, to search engine history to smart camera monitoring, the process of uploading these personal private data to the data center contains the risk of data leakage, so the improvement of security will be what DCC needs to achieve main performance.

3. Intellectual property rights are properly resolved

Closely related to security concerns are concerns about proprietary data and intellectual property. In cloud computing, all user data needs to be uploaded to the data center, and some important information regarded as commercial secrets may be obtained through industrial data obtained by high-quality sensors, so a reasonable solution to intellectual property issues is of great importance to the development of DCC big impact.

4. The cost of bandwidth will be greatly reduced

Sensors connected within the system generate large amounts of data, and in these cases, sending all this information to the cloud would take too long and be prohibitively costly, whereas distributed computing enables high-throughput computing while maintaining security , which will greatly reduce bandwidth costs.

5. Enhancement of autonomy

It is precisely because of the delay and elasticity issues that the independent decision-making of edge computing does not depend on the characteristics of the cloud, which has become a decisive advantage in IoT applications. Therefore, in an emergency situation, the DCC platform can simultaneously monitor itself and the process it is executing, and can also program it, so that it can fully realize the characteristics of decentralization while ensuring its own security and stability.

6. Technical architecture will be standardized

Any major technological breakthroughs belong to a competitive architecture in the early days. Now there are many experiments and solutions in DCC, and certain industry standards will gradually emerge in the next 5-10 years, which will also bring about the rapid development of the entire industry.

7. Blockchain and cloud computing will achieve limited integration

Blockchain technology can indeed create a completely secure and democratic network in theory, but the price that users are willing to pay for "safety" is limited; the integration of blockchain and cloud computing in the future is inevitable, and small node clouds will appear The blockchainization of important nodes, and even limited backup scenarios.

1. This report is an original work of Jingzhun (ID: rong36kr), a professional data research and analysis organization [Jingzhun Research Institute], which is protected by the "Copyright Law" and enjoys the right of compilation and annotation according to law;

【Reprint Notice】

1. This report is an original work of Jingzhun (ID: rong36kr), a professional data research and analysis organization [Jingzhun Research Institute], which is protected by the "Copyright Law" and enjoys the right of compilation and annotation according to law;

3. Commercial reprinting and secondary editing and reprinting are prohibited.

3. Commercial reprinting and secondary editing and reprinting are prohibited.