This year is alibaba’s 13th Singles Day. On Singles’ Day this year, Alibaba pushed its peak computing costs to drop by 50% compared with last year; As of Monday, the xiaoman Donkey unmanned vehicle has delivered more than 1 million packages during the Singles Day holiday. More self-developed technologies of Ali are invested in Double 11. While the core technologies such as chips and servers are tackling difficulties, the technologies are also promoting the cost reduction and efficiency increase of the whole link. In addition to deep cultivation of self-reliance and self-improvement of high-tech, Ali is also practicing the good technology of “everyone benefits, responsibility, open and shared”, hoping to use technology to create better value for customers, to create greater contributions to the society. From higher to better, what’s different about Alibaba’s technology this Year?

The author | Alibaba Tech source | ali technology to the public

This year marks alibaba’s 13th Singles Day.

On Singles’ Day this year, Alibaba pushed its peak computing costs to drop by 50% compared with last year;

As of Monday, the xiaoman Donkey unmanned vehicle has delivered more than 1 million packages during the Singles Day holiday.

More self-developed technologies of Ali are invested in Double 11. While the core technologies such as chips and servers are tackling difficulties, the technologies are also promoting the cost reduction and efficiency increase of the whole link.

In addition to deep cultivation of self-reliance and self-improvement of high-tech, Ali is also practicing the good technology of “everyone benefits, responsibility, open and shared”, hoping to use technology to create better value for customers, to create greater contributions to the society.

From higher to better, what’s different about Alibaba’s technology this Year?

Green low carbon

1. Green technology helps to reduce carbon emissions per order

The carbon cost behind the transaction mainly consists of three parts: first, the carbon cost of commodity/digital transaction itself; Second, the carbon cost in logistics and transportation; The third is the carbon cost in commodity production and consumption.

Low-carbon double 11, must answer three questions: first, how to make the data center more “green”, more energy saving; Secondly, how to improve the efficiency of unified resource scheduling through computing power sharing; Thirdly, how to maximize the efficiency and value of unit force.

In terms of green data center, Ali Cloud integrates AI applications such as low-carbon site selection, clean energy and liquid cooling technology to accelerate green and energy-saving technology innovation and iterative upgrading, and achieve “less electricity, better electricity and green electricity”. In terms of green computing, through technological innovation, resource utilization rate has been significantly increased to help save energy and reduce emissions. This year, the carbon emissions of each order in the trading process have been further reduced. In terms of green computing power, cloud provides a technical framework and economic model for full reuse and efficient utilization of computing power, making computing power itself greener. In terms of digital green supply chain, drive to algorithm, using predictive + decision-making algorithm system, from the evolution of machine learning to deep learning technology stack, integrates data-driven forecasting decision-making optimization, continuous optimization and innovation of the supply chain programs and transported with, reduce damage, improve transport efficiency, help businesses to reduce carbon emissions.

2 technology-driven construction of Alibaba’s green digital supply chain

On November 11 this year, the digital supply chain adopts the algorithm system of prediction + decision, realizing the integrated optimization of data-driven prediction and decision, reducing cargo loss, unified scheduling and reducing waste. In the transportation and distribution scenario, improve vehicle loading rate and transportation efficiency; Through the construction of a inventory, full link can be sold to help merchants share inventory, more accurate replenishment stock, help merchants multi-end sales, reduce unsalable.

3 ali Cloud data center uses green electricity on a large scale to provide green “backing” for “Double 11 Carbon Reduction Plan”.

In order to facilitate the low-carbon and green development of data centers, Alibaba has actively responded to the requirements of the national “dual carbon” strategy in recent years by conducting clean energy and electricity trading. In 2018, Zhangbei Data Center joined zhangjiakou “Four-party Cooperation mechanism” wind power transaction and took the lead in non-water renewable energy power transaction in the national data center industry. In September this year, as one of the first national green power transaction subjects, the company reached a 100 million KWH photovoltaic power transaction and became the largest buyer of green power in the Internet industry in this transaction, practicing the concept of green development and exploring the path to obtain green power through new trading varieties. During double 11, alibaba’s Zhangbei data center is expected to use nearly 30 million kilowatt-hours of green electricity and reduce 26,000 tons of carbon dioxide emissions. From January to September 2021, Alibaba traded 248 million kilowatt-hours of green electricity, according to public data. From 2018 to September 2021, Zhangbei Data Center alone has traded about 600 million kilowatt-hours of new energy power and achieved a total carbon dioxide emission reduction of nearly 523,000 tons.

It is worth mentioning that Alibaba Zhangbei Data Center became the first carbon inclusive pilot project in the industry and was awarded the National Green Data Center in 2020 and selected as a typical low-carbon green case of the Ministry of Ecology and Environment in 2021.

4 Unified scheduling and resource pools support Double 11 in a stable, efficient, and low-cost manner

On the application side, aiming at high concurrency scenarios, Ali technology greatly optimizes transaction links to deal with pulse-type traffic while ensuring consumption experience. On the software side, the unified resource pool is used for unified scheduling to support large-scale off-line mixing and achieve priority scheduling of online services. In terms of cloud, the utilization efficiency of server resources has been greatly improved based on the DpCA server and PolarDB database developed by Ali, which can build a container cluster with a scale of millions and release the technical dividend of cloud origin on a large scale.

5. Green logistics has been comprehensively upgraded. Through intelligent box cutting, green returning and carbon billing, the green mind can fully reach consumers and create a green Double 11 that everyone can participate in

By optimizing carton models and recommending reasonable packing schemes, cainiao intelligent box cutting algorithm supports cainiao packaging materials recommendation in the whole industry, reducing the use of packaging materials by 15% on average, and reducing the weight of 530 million packages in cainiao warehouse a year.

Through the calculation of carbon emissions and the information inspection of all links of green logistics, the first “personal carbon reduction bill” in the logistics industry was launched during tmall Double 11 in 2021, and the “express packaging recycling for eggs” project was launched in 60,000 Cainiao stations, and the recycling mail project of used packaging was piloted in 10,000 Cainiao stations. Open up the loop of express packaging recycling that consumers can perceive and participate in.

Technology innovation

1 Flat brother chip with light 800 scale support search recommendation and other scenarios

During the period of 2021 Double 11, Hanguang 800 supported Taobao search, recommendation and other businesses through Ali Cloud platform, and 100% of THE AI computing power of Taobao’s main search was provided by this chip. The actual application shows that 800 with light effectively gives full play to the advantages of the integration of chip and cloud computing environment, which not only improves the system performance but also reduces the overall energy consumption. Taking the search scenario as an example, compared with the traditional GPU, the algorithm running with 800 with light can increase the efficiency by nearly two times and reduce the energy consumption per unit calculation force by 58%.

2 self-research ibanjiu server new generation hardware technology launched

Iwaki server’s new generation of hardware and technology is applied to Double 11, providing industry-leading computing and storage productivity, and increasing scenario-based performance by more than 30%. This year’s Cloud conference, Alibaba released from research ibanjiu server series, including high performance computing series, high performance storage series, large capacity storage series. The unified fangsheng architecture is used to realize technology and architecture innovation and self-research from chips, components to the whole machine. It uses flexible modularization and full-stack software and hardware integration technology to provide a variety of general and heterogeneous computing and multi-level storage solutions, bringing customers industry-leading, higher reliability and more extreme performance experience.

3 Double 11 Cloud native technology to achieve self-research, open source, commercialization trinity

The middleware evolves towards the product trinity architecture, with the backend BaaS, runtime Mesh and service Serverless, so as to achieve out-of-the-box, pay-as-you-go, and comprehensively improve the efficiency of R&D operation and maintenance. Among them, basic middleware is upgraded to BaaS, microservice is upgraded to MSE, message is upgraded to MQ, observables are upgraded to ARMS, high availability is upgraded to AHAS, full-link pressure measurement is upgraded to PTS, and public cloud products are upgraded. With extreme flexibility, minute-level station construction and capacity expansion are greatly promoted, greatly improving station construction efficiency. Mesh is implemented on a large scale in internationalization, sinking non-business logic to Sidecar and making business focus on business development. The business is Serverless, and FC function computing cloud products are 100% installed in the front end of Taobao shopping guide to achieve rapid capacity expansion and quick response, and help improve the operation and maintenance efficiency of business. In the future, it is expected that more enterprises will adopt Ali Double Eleven middleware to support business peak and quickly share the technology and ability of middleware precipitation.

4 Multi-mode large model M6 supports more than 40 business scenarios of Double 11 for the first time

Superlarge scale models have become the focus of artificial intelligence. Recently, Ali Dharma Institute has realized the world’s first 10 trillion parameter multi-mode large model M6, and released an external platform, covering various single mode and cross mode understanding and generation tasks. As the first multi-mode large-scale model to be commercialized in China, M6 has been applied in more than 40 business scenarios of Alibaba, with daily usage exceeding 500 million.

This year, the grand Model supports Double 11 for the first time. Relying on multi-modal understanding ability, M6 is improving the search and content cognition accuracy of Taobao, Alipay and other platforms. With fluent writing ability, M6 is writing scripts and copywriting for tmall virtual anchor. M6 has been working in Rhinoceros Wisdom Manufacturing co., LTD. Its high-definition image generation ability has improved the work efficiency of designers by 5 times. The development cycle of new clothes has been reduced from 1-3 months to 2 weeks at the fastest.

In particular, IT is worth mentioning that M6 achieves the ultimate low carbon and high efficiency in the industry. Compared with gpT-3, the large model released last year, M6 achieves the same parameter scale with only 1% of its energy consumption, which greatly reduces the threshold for the realization of large model and promotes the development of PRATT & Whitney AI.

5 Large-Scale application of IPv6 on Double 11

In the past three years, Alibaba has actively promoted the large-scale deployment and application of IPv6 in the technology ecosystem. By October 2021, the group’s monthly live IPv6 users have exceeded 600 million. The mobile Internet IPv6 traffic of e-commerce apps Taobao, Tmall, Xianyu and Kaola accounted for more than 76 percent during Singles day. Taobao’s core business function “transaction order” link has been upgraded to IPv6 and one-third of the traffic is carried by IPv6, supporting the Double 11 promotion for the first time. In addition, Dingding has completed the transformation of IPv6 full link, which is the first end-to-end full link application in China to support ipv6-only.

6. Technological innovation of full-stack digital supply chain helps improve business efficiency

Digital supply chain provides end-to-end digital supply chain solutions for the industry and merchants, realizing the coverage of full-link scenarios from procurement to circulation and from merchants to consumers.

This year, the Digital supply chain of Double 11 continues to innovate in technology. We have developed Falcon, a deep timing prediction algorithm, to help merchants to carry out intelligent sales prediction and replenishment and warehousing, and created DChain Forecast, an intelligent timing prediction robot product, so that merchants can also have high-accuracy timing prediction ability. Continue to build the supply chain simulation platform, realize the full-link high-fidelity simulation of the supply chain, support the sandbox simulation of warehouse deployment and network planning, optimize inventory management, and improve the shelf; Provide intelligent performance decision making ability, realize the coverage of warehouse routing, order collection and batch, distribution scheduling and other operational scenarios, ensure the realization of service objectives of different time requirements, such as near and far field, fresh standard products.

The digital supply chain has driven the flow of over hundreds of millions of goods and orders, using digital technology to boost business efficiency and economic vitality.

7 Alinje products fully embrace cloud native, performance and availability of comprehensive improvement

Alidinger fully embraces cloud native, including MaxCompute online unified scheduling, real-time computing Flink version VVP platform upgrade Serverless and Hologres are fully online to cloud native platform, which improves cluster stability and resource elasticity. The MaxCompute unified scheduling cluster is responsible for 40% of the transaction volume, and the performance is also the highest in the mixed cluster. Real-time computing Flink version engine optimization, Hologres high availability capacity building, and other help core business marketing activities analysis, business staff to save resources at the same time, availability is also greatly improved.

8. The cloud native architecture is fully internationalized, and further practices the cloud native value based on business architecture innovation

During Double 11, all core applications were internationalized with 100% cloud native, 100%ASI enabled the ability of Mesh data surface, and the sink of Mesh control surface covered most core systems of core trading link. The cloud native decentralized gateway architecture was successfully implemented in AE scale, receiving large traffic boost and withstood the test of peak traffic on Double 11.

Service architecture: In order to improve the continuous evolution capability of the architecture, the gateway components are containerized and deployed based on container layout to realize the decentralized gateway architecture, setting a new benchmark for the evolution of cloud native business architecture.

Infrastructure: Based on the service grid, enterprise-level traffic governance standards and product architectures are realized, supporting the traffic control and governance of thousands of applications at the service level during the Double Eleven. The implementation of the online traffic governance rules has been upgraded from the application dimension in days to the effectiveness of the online traffic governance rules in seconds, and the per capita efficiency has been increased by more than one time.

Experience to upgrade

1 taobao 3D immersive live broadcast room and virtual anchor to create a new immersive consumption experience

During singles’ Day, Taobao Live will bring consumers a new immersive consumption experience by integrating 3D immersive live broadcast rooms, 3D products and virtual anchors (AI digital people).

Taobao live 3 d immersive studio based on the reality of the extended level the world’s first film and television shooting techniques of cloud, breaking the previous film grade XR filmed live in cloud rendering, AI and complex lighting simulation key technology such as 3 d asset library, cloud, anchor support using ordinary green screen can only live virtual out thousands of square meters of the venue, nearly thousand lamp lighting, special cameras live effect, Become the first commercial self-broadcast film and television shooting solution with low cost, reusable, low carbon and environmental protection advantages.

Live at the same time, taobao not only use advanced 3 d modeling technology, the mass line of goods online to realize 3 d rendering, more in the big data and multimodal technology support, let all the virtual anchor according to the businessman in studio script multimodal commodity introduction, graphic and appearance of products in the process of live, interactive dialogue, the interpretation of problems to solve, Bring consumers new interactive experience of digital space consumption.

2. Multi-language technology innovation promotes sales of cross-border e-commerce merchants

Creative translation technology integrates rewriting and translation capabilities to alleviate the impact of low-quality source language product information and significantly improve user language experience. Currently, AliExpress and ICBU have supported the internationalization of tens of millions of cross-border product information in more than ten countries. Image translation technology automatically locates and identifies the translation area, automatically converts Chinese pictures into English pictures, and easily internationalizes creative ideas. Lazada has realized the internationalization of millions of commodity picture information. Image translation and editing tools have enabled thousands of cross-border merchants. Multilingual generation technology can also integrate fragmented, scattered information, generate attractive material, to provide consumers with timely and rich information.

3. Double 11 will be more open and secure

Business Center Upgraded through center center capabilities to provide a better shopping experience for consumers and merchants.

On the consumer side, the details of the consumer purchase link are constantly polished to improve the purchase experience. With the support of associative word recommendation and adaptive index selection, users can quickly search for the purchased target goods, add good goods and share them in shopping cart with one click, and place combined orders for various goods in live broadcast room. At the same time, through the opening of e-commerce and payment capacity, the business center helps to open up multi-payment and multi-preferential channels, shorten the overall payment path of users, reduce the transaction duration of 5.5s, and improve users’ payment experience.

On the merchant side, through the intelligent algorithm in Taiwan, including commodity inventory warning, advance attention to top goods and explosive products, 0 second refund and other capabilities, to help the merchant efficiently and reasonably stock up; At the same time, for the rights and interests of the broadcast room, promote the single conversion rate up to 6%, the record of the broadcast room anchor to climb the peak.

4. Initiate cross-border B2B live streaming to optimize audio and video experience

In the second half of 2021, the foreign trade industry faces multiple uncertainties due to repeated COVID-19 outbreaks, offline exhibitions and cross-border logistics disruptions, and fluctuations in exchange rates and upstream raw material prices. Alibaba’s international site was the first cross-border B2B live broadcast last year, and has hosted more than 300,000 live broadcasts so far.

In order to provide customers with better cross-border live broadcasting and customer service experience, XG Lab of Dharma Institute cooperates with video cloud, enterprise intelligence and other teams to realize extremely weak network countermeasures through cross-layer fusion of audio and video applications and network transmission optimization, optimize qOE-oriented bandwidth estimation and anti-packet loss technology, and achieve 70% packet loss countermeasures. Through more than 20 transmission technologies such as bandwidth estimation, SVC and anti-packet loss, the audio and video experience under various weak network scenarios is guaranteed. The core experience indicators such as audio and video sharpness and lag rate are improved, and the video lag rate is reduced by 80%, especially in extreme weak network scenarios.

5. AnalyticDB of cloud native data warehouse: greatly improve the real-time performance of query analysis

Ali Cloud database has been stably supporting Double 11 for many years, undergoing quenching in extreme flow scenarios. In addition to ensuring the stable and smooth basic disk, during the promotion period of this year, the database has greatly improved user experience through comprehensive cloud original biochemical, so that technology can help businesses produce more valuable consumer experience, and continue to empower users through technological innovation, leading the technology development path.

This year, Ali Group through the introduction of Ali cloud native relational database PolarDB+ cloud native data warehouse AnalyticDB (ADB) + one-stop online data management platform DMS, enabling trading business in Taiwan to achieve convenient search for full orders, to achieve personalized association word recommendation, so that the retrieval efficiency is greatly improved. Effectively improve the search experience of mobile users.

In addition, AnalyticDB has added hundreds of system core indicators this year. With abundant tools and the ability of continuous precipitation, risks are discovered in advance to ensure the silk-like smoothness of double 11, and no delay in writing peak million-level TPS, while greatly reducing the workload of engineers.

PolarDB+AnalyticDB+DMS also supports the core business scenarios of Double 11, such as real-time buyer and seller database synchronization and historical order search.

6. Cloud native in-memory database Tair: Ensure that the shopping cart displays the user’s price in real time

Tair, the native in-memory database of Ali Cloud, as a sharp instrument to promote the peak load of traffic on Double 11 in the past years, not only maintains the smooth delay of sub-millisecond level under the peak value of billions of QPS, but also makes technological innovation in the core experience scene of e-commerce transactions.

Through the native persistent memory storage structure, Tair assisted shopping cart scenario realizes high concurrent traffic, real-time display of dynamic offers and coupons after the purchase price, so that users can see the actual price of goods at a glance, significantly improving consumer purchasing experience.

It is reported that Tair, the in-memory database developed by Ali, was born in 2010. It is a cloud native in-memory database that supports high concurrency and low latency access. It is fully compatible with Redis and has passed the test of Double 11 promotion for many years, providing core online access acceleration ability and significantly improving system throughput.

Technology is suitable for the elderly, so that digital services do not create digital divide

At present, 12 Alibaba apps have joined the barrier-free action, so that everyone can equally enjoy the benefits of technology. Taobao “elder mode”, through the introduction of font size adaptation, personalized recommendation, ASR voice conversion and other technologies, eliminates the digital divide, helping elderly friends to read clearly, understand, and feel assured;

Taocai provides “silver hair version” for the elderly group, with larger font size and easier operation. Meanwhile, it optimizes the user experience of low-end machines, increasing the opening speed of small programs by 50% and increasing the performance of purchase by 30%.

Ele. me continues to optimize the user experience of elderly users and people with disabilities, covering the whole process of shopping guide, search and ordering. In the process of developing the elder mode, the font size and the function modules commonly used by the silver-haired people were optimized. In addition, Ele. me has launched a barrier-free communication system for silent Blue Knights and matched and received orders for disabled riders through an intelligent dispatching system. In the past year, the intelligent dispatching system reduced the distance of each rider by 7%, enabling hundreds of thousands of disabled and silver-haired users to better enjoy local life services.

Industrial smart

AI image generation technology helps fast reverse fashion design

In the fashion industry, how to efficiently and accurately mass produce innovative design styles in line with the market trend and shorten the cultivation cycle of new models and inventory risk has always been a pain point for merchants.

Image generated by AI technology, ali has developed a set of intelligent trend design solutions, according to the trend, high-resolution real-time output design style, support different levels of editing operation, and through the personalized distribution reduce brand homogeneity, and improve the new product research and development speed and precision, make the brand and the consumer demand docking more smoothly.

In this year’s “Double 11” app, thousands of AI-designed products were launched and applied by shoe vendors.

2 Small man donkey unmanned vehicle to achieve large-scale delivery

This year, more than 11,350 Ali donkeys and Cainiao Courier stations will drive into colleges and universities across the country, delivering more than 50,000 packages of express delivery every day, and it is expected to deliver more than 1 million packages. “Little Man Donkey” combines the autonomous driving technology with cainiao terminal logistics and instant delivery scenarios, realizing the large-scale implementation of product-driven autonomous driving. In the logistics industry, the end logistics has always been the link with the highest cost and lowest efficiency in the whole logistics system. Unmanned vehicles drive in the unstructured end scene, which needs to identify and deal with a large number of traffic participants, such as motor vehicles, non-motor vehicles, pedestrians, animals, and other emergencies. One million orders are supported by the algorithm technology, operation technology and security capability of The Little Man Donkey. Taking the algorithm as an example, the little man Donkey currently identifies about 40 million obstacles per bike every day and handles about 5,000 interactions with other traffic participants every day.

The original link

This article is the original content of Aliyun and shall not be reproduced without permission.