Moving Towards an Intelligent Era, Redefining Data Infrastructure
On May 15, 2019, Huawei Executive Director of the Board and President of ICT Strategy & Marketing Wang Tao delivered a keynote speech titled Moving Towards an Intelligent Era, Redefining Data Infrastructure at Huawei's Database and Storage Product Launch. The speech script is as follows.
Ladies and gentlemen,
It's great to see you today. It's my honor to talk with you again. On behalf of Huawei, I would like to extend the warmest welcome to everyone here at the Launch!
In 2017, Huawei released its new vision and mission – to bring digital to very person, home and organization for a fully connected, intelligent world. We are committed to working with partners to build a fully connected, intelligent world, and data infrastructure is the core driver behind an intelligent world. Today, I would like to discuss how we move towards an intelligent era and redefine data infrastructure.
Society is quickly moving from an agricultural and industrial economy to a digital economy. According to Huawei's Global Industry Vision, global data will increase from 32.5 ZB in 2018 to 180 ZB in 2025. A massive amount of data has posed huge challenges and also created unprecedented opportunities. The wide application of AI helps identify the potential value of data and significantly improves the competitiveness of businesses in the digital economy. According to the analysis of Oxford Economics, the growth rate of the digital economy is 2.5 times that of the global economy; the ROI of the digital economy is 6.7 times that of the non-digital economy; and the digital economy will be worth 23 trillion US dollars in 2025. The digital economy has become a new engine that drives the growth of the global economy. Data is becoming a new means of production, and intelligence is becoming a new form of productivity.
All industries are accelerating their process of going intelligent. More and more companies are realizing that data infrastructure is the key to the success of an intelligent transformation. Data infrastructure can realize data computing, storage, and analysis by utilizing computing resources, storage resources, and data processing platforms respectively. Let's take a look at some industry case studies to learn how "Data + Intelligence" will transform industries and create new industry and social value.
"Data + Intelligence" will transform all industries
"Data + Intelligence" is supporting efficient processing of a massive amount of data in the autonomous driving industry. For example, in level-5 autonomous driving, each car will generate 64 TB of data every day. From development to mass production, over 10 billion kilometers of autonomous driving experience need to be gained, and this will generate over 50 EB of data. Octopus, the autonomous driving cloud service facilitated by HUAWEI CLOUD, supports simulation tests across 30,000 virtual scenarios. This helps increase model training efficiency by 80%, and significantly shortens the time for development from months to weeks.
"Data + Intelligence" is also applied in financial security. Traditional financial risk control platforms cannot satisfy the business needs of completing risk control computing within 100 milliseconds. Huawei's real-time financial risk control solution, powered by big data, in-memory computing, and AI, can complete risk control computing of tens of millions of data within 30 milliseconds. In 2018, this solution helped the China Merchant Bank reduce transaction loss of over 50 million US dollars, and the financial risk rate dropped by 83%.
I would like to share with you another fascinating example. In 2018, Qinghai, a province in western China, achieved a world record of using clean energy, such as hydropower, solar power, and wind energy, for power supply for nine consecutive days. This is realized through accurate matching of power supply and market demand. Huawei worked with Qinghai Power Company to build a big data center. The accuracy of power generation and usage forecasts has reached over 97% through the sharing and analysis of the data of power generation, distribution and usage. What's exciting is that Qinghai plans to complete the construction of a power system fully supplied by clean energy by 2030, with the aim of providing power through clean energy all year round.
In addition, "Data + Intelligence" is also protecting our ecosystem. The company Rainforests monitors an area of 2,500 square meters of tropical rainforests in Costa Rica with equipment powered by solar energy. With HUAWEI CLOUD's capabilities of storing and analyzing a massive amount of data, the monitoring equipment is able to process the complex audio records of the rainforests in real time, and accurately identify noises generated by electric saws and trucks to stop illegal logging. Now let's see a video about it.
Redefining data infrastructure in the three areas of computing, storage, and data processing
Data infrastructure deployment is a must for industries to accelerate intelligent transformation. Currently, only less than 2% of enterprise data is stored, of which no more than 10% is analyzed and utilized. Much value of data is still unreleased. The diversity of business and data is driving the diversity of computing. Single computing architecture cannot process all types of data across all scenarios. For example, data models such as CPU-enabled image processing and neural networks consume a lot of energy yet only release less than 20% of computing power of the system. Using a heterogeneous platform that consists of multiple computing architectures is the key to addressing the scarcity and high cost of computing power. Over the past ten years, with the fast development of enterprise applications, IT systems have been data silos. This blocks the sharing of data and flows of resources. The management and extension of IT systems are complicated, with low storage efficiency. Traditional data analytics platforms cannot be powered by AI, and does not support real-time stream processing, so the value of data is not fully unleashed.
Companies need faster and more intelligent data infrastructure in the intelligent era. Huawei has redefined data infrastructure in the three areas of computing, storage, and data processing, helping industries to move faster into an intelligent world. In computing, Huawei is committed to leading diversified computing and driving the development of diversified computing architectures for more abundant and cost-effective computing power. To tackle low storage efficiency and complex management, Huawei has redefined the storage architecture, delivering a huge boost in efficiency, and leading the transformation towards intelligent storage. In response to needs for more intelligent, real-time requirement analysis, Huawei has redefined data processing platforms, enabling more intelligent analysis and creating greater value from data.
First, let's take a look at how Huawei is leading diversified computing. Huawei is driving the shift from a homogenous computing architecture based on x86 + GPU to a heterogeneous computing architecture based on x86 + GPU + ARM64 + NPU. Huawei has introduced AI management and intelligent acceleration capabilities based on the x86 architecture and launched the FusionServer Pro intelligent servers. Huawei has also developed the Taishan series servers based on the ARM64 architecture, with the best performance in the industry. The Atlas intelligent computing platform, powered by the Ascend chips, is the first in the industry that coordinates between devices, edge, and cloud.
Redefining the data processing platform for more intelligent analysis and faster value creation from data
Huawei will continue to make strategic investments in data processing, innovate, and make breakthroughs. Traditional companies mainly process structured data, with "Database + BI applications" as the dominant model. In the intelligent era, more diversified data will be processed, including structured and unstructured data. Powered by cloud and distributed technology, the modern data processing platform that combines database, big data, and AI is becoming mainstream.
Huawei has redefined the data processing platform in three major aspects: heterogeneity, intelligence, and synergy. Heterogeneity means that Huawei databases run on the computing power from diversified sources, including x86, ARM, GPU, and NPU, for best performance. Intelligence means that AI is used to significantly increase automation in database management and optimization. Synergy means that the three functional modules of big data, database, and AI processing work together for coordinated computing and converged analysis of diversified data, delivering maximum value from enterprise data.
Huawei provides lifecycle, full-stack big data platform services. HUAWEI CLOUD has released 13 database services under four categories based on widely-accepted database ecosystems. Huawei has also released 59 services based on full-stack, full-scenario AI solutions, with more than 160 functions, enabling inclusive AI that is affordable, effective, and trusted. Huawei released the ModelArts, a one-stop AI development platform that lowers the bar for the use of AI. According to the DAWNBench ranking released by Stanford University, ModelArts ranked No.1 in both AI training and inference. Huawei will continue to make investments in R&D and innovation in data processing platforms and provide customers with a wider array of database products and services.
GaussDB, the world's first AI-native database
Today, on behalf of Huawei, I would like to launch the world's first AI-native database, GaussDB. It is also the industry's first enterprise-class database compatible with ARM processors.
GaussDB delivers two disruptive breakthroughs. First, it introduces AI technology throughout the lifecycle of the database for the first time, enabling self-O&M, self-management, self-tuning, and self-fault-diagnosis. Based on the optimality theory, GaussDB offers a 60% performance improvement in transactions, analyses, and hybrid load by applying self-tuning deep reinforcement learning algorithms. Second, GaussDB supports heterogeneous computing and runs on the computing power from diversified sources, including the x86 processor, ARM processor, GPU, and NPU, for best performance. Huawei's GaussDB ranks No. 1 in performance in a TPC-DS test. In addition, GaussDB supports a variety of scenarios, such as on-premise deployment, private cloud, and public cloud. Deployed on HUAWEI CLOUD, GaussDB provides full-featured high-performance data warehouse services to customers in the financial, Internet, logistics, education, and automotive industries. My colleague will provide you with more details on GaussDB.
FusionStorage 8.0, a next-generation intelligent, distributed storage system
Storage systems are also an important part of data infrastructure. On behalf of Huawei, I would like to launch the second product: FusionStorage 8.0, a next-generation intelligent, distributed storage system.
Nowadays, we face three major challenges: dramatic increase of data traffic, a wide variety of data types, and real-time data analysis. Addressing these challenges requires larger-scale storage capacity, faster business response, and intelligent O&M management. FusionStorage 8.0 redefines the storage architecture and enables more intelligent, efficient storage by implementing "Storage for AI" and "AI in Storage".
First, "Storage for AI" helps eliminate the need to copy data multiple times through data convergence and sharing, making AI-enabled data analytics more efficient. FusionStorage 8.0 supports block, file, object, and HDFS storage services, breaking down data silos in the storage system. This resolves problems related to the sharing of diversified data. In addition, FusionStorage 8.0 makes the sharing of data sources possible through multiple protocols. This makes data migration and analytics more efficient and prevents meaningless data migration. Second, "AI in Storage" means applying AI in storage product lifecycle management, enabling intelligent O&M in terms of resource planning, service provisioning, system tuning, risk predictions, and fault location.
Huawei has redefined data infrastructure in the three areas of computing, storage, and data processing. In computing, Huawei is committed to leading diversified computing and enabling more abundant and cost-effective computing power. In the storage area, Huawei has redefined the storage architecture, delivering a huge boost in efficiency and driving the transformation towards intelligent storage. Huawei has also redefined data processing platforms, enabling more intelligent analysis and creating greater value from data. Innovation is the core element of Huawei's product research and development. We will continue to leverage our advantage in "Connection + Computing + Cloud" synergy to lead the development of the ICT industry and build an engine for the intelligent era!
Thank you for your time! I hope you will have a good time in Beijing!