For enterprises, digital transformation is a "passport". IDC believes that what enterprises urgently need to solve at this stage is to improve their digital capabilities, including: the ability to deeply integrate with business; Data processing and mining capabilities; And IT technology operation management ability. Especially the ability of data processing and mining, because digital transformation promotes enterprises to change from process-oriented to data-oriented, the ability of processing and mining massive, heterogeneous and multi-type data is the premise of releasing data value, and the control of data life cycle is the guarantee of releasing data value. With the introduction of a large number of new technologies in digital transformation, the complexity of IT has become higher, and the ability of enterprise IT technology operation and management is the key to enhance enterprise IT productivity.
To solve the problem of digital transformation, we need a digital platform with three characteristics: integration, intelligence and inheritance. This is the viewpoint put forward by Huawei and IDC in the white paper "Embrace change, outsmart the future-digital transformation of enterprises with broken digital platforms" jointly released in March 20 19. Integration mainly refers to the combination of traditional technology and innovative technology; Intelligence mainly refers to platform intelligence and intelligent capability output; Inheritable architecture mainly refers to the architecture created through the concepts of decoupling, functional reuse and configurability. Carrying these three viewpoints is a new generation of distributed enterprise-level technology.
On May 5th, 20 19, Huawei released GaussDB, the industry's first intelligent distributed database supporting ARM architecture, and FusionStorage 8.0, a distributed storage. As a new generation of data infrastructure, Huawei interprets a digital platform with three characteristics: integration, intelligence and inheritance. Wang Tao, managing director of Huawei and president of ICT strategy and marketing, said at the press conference that thousands of industries are accelerating the process of intelligence, and more and more enterprises have realized that data infrastructure is the key to the success of intelligence. Huawei redefines the data infrastructure around three areas: computing, storage and data processing, and accelerates to the intelligent age.
Behind the concepts of cloud and industrial internet discussed today is the arrival of a new era and the great migration of architecture. The traditional enterprise-level technology is to ensure the high reliability, high availability and high stability of data storage, calling and other operations in a single application and single machine environment, especially to meet the ACID (atomicity, consistency, isolation and persistence) requirements of financial-level transaction processing, and to provide data management support for key businesses of enterprises. With the migration of enterprise technology to cloud architecture, database technology is also facing transformation.
20 18 distributed database based on cloud computing technology has become a hot spot in the industry. Simply understood, cloud computing technology is to replace the "stand-alone" environment with a distributed computing environment composed of X86 server clusters. The computing tasks that were originally completed by several minicomputers have to be distributed to hundreds or even thousands of X86 servers, and may also be operated across data centers. The challenge can be imagined. Especially for financial services such as online payment, you can't make mistakes when the network is disconnected or there is a problem with network connection, and you can't affect the user experience because of slow response.
2065438+August 2008, China Payment and Clearing Association and China Information and Communication Research Institute jointly held a seminar on financial distributed transaction database, and discussed the road of distributed transformation of core database with industry manufacturers and users, and released the White Paper on Financial Distributed Transaction Database. The popularization of financial distributed transaction database provides industrial support for distributed database to enter key business systems of enterprises. As an enterprise ICT solution provider, Huawei began to develop a data warehouse for big data analysis as early as 20 12. Based on the strong consistency between the traditional relational database SQL engine and transactions, it reformed the distributed and parallel computing, and it took six years to build a distributed database for PB-level massive data analysis.
In addition to OLAP data warehouse, Huawei also cooperates with industry users to develop a distributed transaction database for OLTP. In 20 17, Huawei cooperated with China Merchants Bank to establish a distributed database joint innovation lab, and developed a high-performance OLTP distributed database with enterprise-level kernel, which fully supported distributed things and met the requirements of the financial industry for strong data consistency, with a single processing capacity of one million per minute.
The new GaussDB database products released this time include: online transaction OLTP database, online analytical processing OLAP database, and transaction and analysis hybrid processing HTAP database. Huawei GaussDB database integrates AI technology into database design, development, verification, optimization and operation and maintenance, which can realize self-tuning, self-diagnosis and self-maintenance based on AI, make the database more efficient and intelligent, and lead the development of database architecture.
Furthermore, the GaussDB series database released this time is the first distributed database supporting ARM chips in the industry. Huawei promotes the rapid development of computing architecture from a single computing architecture based on X86+GPU to a heterogeneous computing architecture based on X86+GPU+ARM64+NPU. Based on X86 architecture, Huawei introduced AI management and intelligent acceleration capabilities, and took the lead in launching FusionServer Pro, an intelligent server. Based on ARM64, build Taishan server with the strongest performance in the industry; Atlas intelligent computing based on Ascend chip realizes the first artificial intelligence platform for end-to-end cloud collaboration in the industry. GaussDB can make full use of and integrate various heterogeneous computing power combinations such as ARM, X86, GPU and NPU, and greatly improve the database performance.
Wang Tao emphasized that as the world's first AI native database, GaussDB has two revolutionary breakthroughs: First, artificial intelligence technology is introduced into the whole life cycle process of the database for the first time to realize self-operation, self-management, self-tuning and fault self-diagnosis. In the trading, analysis and mixed load scenarios, based on the optimization theory, the self-adjusting algorithm of deep reinforcement learning was pioneered, which improved the average performance of the industry by 60%. Second, support heterogeneous computing, give full play to the advantages of diversified computing power of X86/ARM/GPU/NPU, and maximize database performance. Huawei GaussDB ranks first in the authoritative standard test set TPC-DS. GaussDB also supports local deployment, private cloud, public cloud and other scenarios.
In the distributed computing environment represented by cloud computing, data management solution not only needs distributed database, but also has become the main framework of distributed database design, in order to better expand and reduce capacity and meet diverse data storage needs. Distributed cloud architecture is to support the architectural design requirements such as computing, storage separation and multi-tenancy.
GaussDB realizes a highly available, reliable and stable distributed database from the database level. FusionStorage 8.0 released this time is a distributed storage architecture, which innovatively realizes that a system supports block, file, object and HDFS protocols at the same time. 1 set of storage supports four types of storage capabilities, which is suitable for mixed loads of all business scenarios, and finally makes "one set of storage in one data center" possible.
IDC's Quarterly Tracking Report on China Software-Defined Storage (SDS) and Hyperconverged Storage (HCI) System Market 20 18 Fourth Quarter shows that the software-defined storage market achieved a year-on-year growth of 54.9% in 20 18. The market share of software-defined storage in China increased steadily, reaching 22. 1% respectively. Huawei's file solutions have been recognized by the government, radio and television, telecommunications and other industries, and China ranked first in the software-defined storage market in 20 18.
FusionStorage 8.0 uses Huawei's arm-based processor Kunpeng 920 for acceleration, which improves IOPS by 20%. Combined with Huawei's AI Fabric lossless network, the delay is further reduced by 15%. Based on Huawei's chip and algorithm accumulation in the fields of computing, network and storage for many years, FusionStorage 8.0 achieved the single-node performance of 168000 IOPS and the delay of 1ms in the performance test of SPC- 1, and became a new choice for carrying key applications of enterprises.
In addition, through the cloud training of Huawei Cloud and local AI chips, FusionStorage 8.0 will intelligently manage the whole life cycle of business use, such as storage resource planning before the business goes online, risk prediction and fault location during use, greatly improving storage efficiency and helping industry customers to cope with new data challenges in the intelligent era.
At the press conference, Wang Tao emphasized that FusionStorage 8.0, a new generation of intelligent distributed storage, has greatly improved the efficiency from the two dimensions of "Storage for AI" and "AI in Storage" by redefining the storage architecture, leading the storage intelligence. First of all, "Storage for AI" makes AI analysis more efficient by integrating * * *. Secondly, "AI in Storage" takes the lead in integrating AI into the whole life cycle management of storage, and realizes intelligent operation and maintenance from the aspects of resource planning, business distribution, system optimization, risk prediction and fault location.
Liaoning Mobile adopts Huawei converged storage. As the largest mobile communication operator in Liaoning Province, Liaoning Mobile has been exploring the application of advanced storage solutions in its own IT system. Due to the rapid development of 5G, the application of Liaoning Mobile's key database is also developing in the direction of cloud, and distributed storage should also meet its requirements of reliability and high performance. After in-depth analysis of the needs of Liaoning Mobile, Huawei first developed and tested distributed storage on a small scale at the edge. After a lot of experiments and tests, the performance and reliability reached expectations, and finally decided to migrate all its services to FusionStorage. The scheme adopts the characteristics of double activity, writable snapshot and end-to-end DIF, and successfully completes the migration of billing, business analysis, B2B and other systems from old storage to integrated storage, helping Liaoning Mobile's storage architecture enter a new historical stage.
It is worth mentioning that Huawei's distributed database and Huawei's distributed storage are deeply integrated, and the operation of the database sinks to the storage node, which greatly improves the performance of the distributed database. Using new network technology and artificial intelligence technology, Huawei helps users to improve the throughput of data centers, improve the scalability of network applications, and can automatically tune.
In addition to launching a new generation of breakthrough distributed database and storage technology, Huawei is also actively working with customers and partners to build an open, cooperative and win-win industrial ecology in the field of database and storage from the aspects of industry applications, platform tools, standards organizations and communities. In terms of industry applications, Huawei has long-term cooperation with independent software developers such as iSoftStone, Shenzhou Information, Donghua Software, Yihualu, UFIDA Government and AsiaInfo International. On the platform and tool level, Huawei and Tableau, Fansoft, ARM, Veritas and other partners jointly innovate. At the level of standards organizations and communities, Huawei has deeply participated in the construction of organizations and communities such as OpenSDS, China Artificial Intelligence Industry Alliance, OCP, OpenStack and CNCF Foundation.
To sum up, the release of Huawei's full-line distributed database and distributed storage products is the latest achievement of Huawei's three characteristics of digital platform: integration, intelligence and inheritance. The combination of Huawei's distributed database and distributed storage can eliminate data islands of various business systems, build data modeling, analysis and value mining capabilities for industry scenarios, aggregate, integrate and analyze multi-source heterogeneous data, form a unified total data and database, and realize data value mining and * * * sharing. AI-based intelligence can efficiently manage infrastructure, empower industry application development and iteration, and comprehensively help enterprises break through the "glory road" of key applications in the cloud. (Text/Ning Chuan)