About us

Quality oriented, customer-oriented, hardworking, pragmatic and innovative

<Return to the public list of news

Huawei Cloud DevCloud: AI and trustworthiness, new thinking of software development

Release time: 2020-09-28 16:18:11

At the end of 2018, the Board of Directors of Huawei issued the Resolution on Starting a Thorough Change, Improving Software Engineering Capabilities, and Creating Credible High Quality Products. It plans to carry out systematic changes in five years, invest $2 billion, comprehensively improve the company's software engineering capabilities, and provide credible ICT high-quality products, services, and solutions to Huawei's customers.


For a while, trustworthiness became the hottest word of the year for Huawei. Since last year, Huawei has launched a comprehensive reform from the definition of trusted products, trusted design, trusted implementation, trusted testing, trusted operation and maintenance, trusted Build In processes, culture and even trusted values.


In the past three decades since Huawei was founded, there have been few such top-level design and R&D changes covering the whole group. The first is the IPD process introduced and established around 1999, the second is the agile change launched around 2009, and the third is the software engineering change launched in 2018 with credibility as the main goal. 2019 is a year when Huawei is facing unprecedented challenges. The company still decided to invest a huge amount of money to launch a group reform involving more than 80000 R&D employees, across consumer business, enterprise business and operator business. It is certain that it is not because Huawei has more money, but because Huawei has captured something keenly behind it?

         

A few days ago, during the HUAWEI CONNECT 2019, at the Huawei Cloud Service Developer Session "Next Generation Software Engineering, All Cloud, More Trusted, Smarter, and More Efficient", the Huawei Cloud preacher shared the "AI Driven Software Development", mainly introducing Huawei Cloud DevCloud's new exploration of software development in AI, and driving software development through AI, Comprehensively improve software engineering capability and practice, and create credible high-quality products. At the 2019 Huawei Full Connectivity Conference, two experts from Huawei Cloud DevCloud first unveiled Huawei's thinking and practical exploration of trusted next-generation software engineering in the past year.


Huawei Cloud DevCloud is a one-stop software development platform that precipitates Huawei's 30 years of R&D thinking and practice, and carries lean, agile, and DevOps best practices. Huawei Cloud DevCloud has helped many customers significantly improve R&D efficiency and quality. The team building DevCloud not only provides services on the cloud to external customers, but also undertakes the delivery of the trusted R&D tool chain of more than 80000 internal R&D staff, and is responsible for Huawei's trusted tool chain, R&D efficiency and experience results. The DevCloud team has always been committed to providing the most advanced R&D equipment to internal and external users.


In the context of the AI era, software developers, as builders of the AI digital world, have been trying to explore how to help users benefit from AI. As a builder of R&D tools, Huawei Cloud DevCloud experts I have been thinking and exploring whether software developers can also benefit from AI. After years of exploration, AI has achieved remarkable results in code review and code repair.


  1. Huawei has been exploring the introduction of AI automatic inspection in code inspection for many years. The latest results are: Nearly 50% of AI's code review opinions are consistent with people's review opinions;


  2. Huawei's exploration of code repair focuses on automatic repair of dangerous functions. The defect repair rate is about 60% the same as that of manual repair, and the actual repair accuracy rate is more than 85%.


The essence of software is the product of an intellectual activity, the accumulation of knowledge and experience, and the highly integrated industry of art and engineering. therefore Huawei Cloud DevCloud experts It is believed that the application of general machine learning in software development still has great challenges, such as difficulty in judgment, low error tolerance, knowledge divergence, etc. In order to deal with these unique challenges in the software industry, the idea of human-computer collaboration should be adopted. Machines should be used to complete repeatable and deterministic work, and people should be used to complete decision-making, creative and uncertain work, and people and machines should cooperate and interact.


The chief product evangelist and senior product manager of Huawei Cloud DevCloud brought Huawei's credible thinking and many practices during 2019HC. There is a deep understanding and practice of "hands full of mud" about why the change is needed and how to promote and carry out the change.



In the sharing of Trusted Software, the Cornerstone of the Internet of Everything Era, Huawei Cloud DevCloud experts systematically reviewed the history of software engineering for nearly 50 years, and divided software engineering into three main stages according to the era problems it solved:


  1. Stage 1 (learn from manufacturing industry, manage deviation, and move from manual workshop to software industrialization). The earliest software development was like a manual workshop. The software development process was often chaotic, disordered or even out of control. The success of software projects even depended on "worship" luck. Successful software projects cannot be guaranteed to still succeed. Under the background at that time, software could only learn from the hardware manufacturing industry nearby. Deviations in the process of managing software, such as CMM, object-oriented, waterfall development mode, were all products of this background. In this phase of software engineering, the quality of software development is guaranteed, controllable and orderly.


  2. Phase 2 (gradually recognize the nature of software and improve efficiency). Although the software engineering in the first stage controlled the deviation of the software to a certain extent, it also inhibited the creativity of the software, resulting in the software development becoming too procedural and mechanized, and the changes being strictly controlled. At that time, the rise of the Internet industry coincided with the large-scale provision of software products to consumers. At this time, demand was more volatile, market competition was fierce, TTM was shorter, and creativity was more needed. Therefore, more and more software practitioners began to call attention to and return to the essence of software: software is an intellectual and mental activity, a combination of art and engineering. The new concept represented by Agility has gradually become popular. The essence of DevOps, which is very popular in recent years, is also the extension of agile and lean thinking to operation, maintenance and operation.


Every stage of software engineering has its own background. Today, the construction of the digital world is hot, and the era of the Internet of Everything has begun. However, when everything blessed by AI is online, information security, environment and personal safety, privacy, reliability, and continued availability have become a growing concern and concern for governments, enterprises, and ordinary users.


In recent news reports, the sudden upgrade of OTA enabled cars has made it impossible for cars to move forward, which is both dangerous and affects traffic. At the same time, we also see that current AI and high-definition photography can reproduce fingerprints taken by scissors, leading to the risk that users' fingerprints will be abused. There are more and more such cases. Whether software products can be trusted has become the core issue of whether software products can be trusted by users.


Because Huawei's business scope is extensive, covering different businesses from ToB to ToC, Huawei is keenly aware that customers will have higher and more stringent requirements for Huawei products in the era of the Internet of Everything. Therefore, on the basis of analyzing the standards and papers of software engineering in the industry, Huawei formally proposed that Huawei should comprehensively reform its software engineering, take credibility as the sole goal, and evolve from result security to result credibility, and further evolve to process credibility.


Based on this reform requirement, Huawei has added privacy to the five basic characteristics of industry credibility, and formally formed six dimensions of Huawei's characteristics for credibility.


In order to guide the reform, Huawei has also released a credible high-level framework.


Huawei Cloud DevCloud experts said in their sharing that it is expected that there will be 100 billion connections by 2025. In the context of the interconnection of machines and people, trusted products, trusted connections, trusted AI, and trusted systems have become the focus and research of the whole society. As the main carrier of the digital world, software trustworthiness is becoming the main research direction of the next generation software engineering.


Huawei Cloud DevCloud experts believe that the third phase of software engineering should focus on solving the problem of software trustworthiness. On the basis of taking quality and efficiency into account, we should make software products that can be trusted from the perspective of process, results and the whole life cycle, reduce the doubts and concerns of the public, enterprises and government about the era of the Internet of Things, and truly achieve a credible digital world.


Huawei attaches great importance to the theoretical basis, but Huawei does not shout slogans, and practice leads to true knowledge. Huawei Cloud DevCloud experts They also shared Huawei's practice in trusted software engineering in the past year, which has pitfalls, pitfalls and gratifying innovations. Such as trusted product definition, build in trusted tool chain, clean code, etc.


Huawei Cloud DevCloud, as the external incubator of Huawei's R&D concepts, practices and projects, will constantly provide external customers with Huawei's internal changes on agility, DevOps and even credibility through cloud services according to customer's scenario needs. I hope to work with customers to meet the challenges of the Internet of Everything era.

 


As a one-stop cloud DevOps platform, Huawei Cloud DevCloud integrates Huawei's nearly 30 years of research and development practices and cutting-edge concepts, provides developers with research and development tool services, and makes software development simple and efficient. It now supports free use for less than 5 people, and can book free product demonstrations and technical exchanges, For details, please click "Read the original text" to view the official website of Huawei Cloud, or Baidu search "DevCloud".



/template/Home/Zkeys/PC/Static