At the 2017 Hangzhou Yunqi Conference, Alibaba Cloud President Hu Xiaoming gave a keynote speech entitled "Intelligent Computing, Born for the Future". Among the many contents, Alibaba Cloud's self-developed database product POLARDB has attracted widespread attention.
Since its birth in the 1960s, databases have permanently changed human's management and cognition of data, and today's big players such as DB2, Oracle, Informix, MySQL, etc. have emerged one after another. Now, from logging on Taobao shopping to social networking sites Big talk is inseparable from the support of the database behind it.
Databases are everywhere.
If you want to explain in detail why Alibaba Cloud chooses the path of self-developed databases, you can probably start from the following three aspects.
The database is already in the era of having to change.
The first large-scale outbreak of Chinese databases was in the telecommunications industry. In 1995, in order to solve the problems of billing and metering, China started the first large-scale database construction. However, in the more than 40-year history of Chinese databases, self-developed databases are mainly concentrated in the government, military industry and other fields of national economy and people's livelihood.
On September 21, 2017, Alibaba Cloud released a new generation of cloud database product POLARDB. This is the first domestically produced general-purpose high-performance self-developed database. For the first time in the high-performance general-purpose database, there is a Chinese face. Previously, this piece has been monopolized by foreign manufacturers.
History has begun to change quietly. Today, with the expansion of massive data, the desire for change in industries such as finance, retail, and logistics indirectly puts forward higher requirements for database technology. Among the many Internet companies in China, why did Alibaba Cloud take the lead in launching a domestic general-purpose high-performance self-developed database?
Back in 2011, Internet companies represented by Alibaba led all walks of life in China to start "Internet +". Traditional IT technology can no longer meet the IT needs after "Internet +", such as: traditional bank IT systems The basis of the design is the number of branches and service windows of the bank. Based on it, the total number of transactions in the bank can be calculated in a day, because to choose the corresponding IT infrastructure, the traditional IT technology is to select the corresponding IT infrastructure. best platform.
But in the Internet era, big promotions and events, and the fluctuation of every large Internet platform will make the transaction volume unpredictable. The Internet technology architecture represented by cloud computing has become the best platform to support this business form.
The emergence of more unicorn companies has once again challenged the traditional database architecture.
Going back to the biggest application scenario of traditional databases: in the financial field, the transaction amount of Yu’ebao has recently exceeded one trillion, and it has become the world’s largest monetary fund. Since its launch in 2013, its business has grown by 14 times in just half a month. It brings a huge challenge to the underlying technical architecture. On the one hand, the time for clearing every night changes from the first half an hour to the next 1 hour, 2 hours, 4 hours... Finally, it takes all night to solve the problem.
From the perspective of traditional IT, the price of such expansion is in the tens of millions, and the cost exceeds the sum of the IT costs of Tianhong Fund over the years.
If this story is put into the whole of China, Chinese enterprises will encounter a database cost crisis in the future. Among Chinese Internet companies, the cost of database has already accounted for 20% to 40% of IT facilities. More importantly, " After the "Prism Gate", more and more people realized that the security of the database cannot be purchased at a high price, and the national interest will always be above everything else.
充分榨干硬件红利突破性能瓶颈
阿里云全新发布的云数据库产品POLARDB首次采用第三代分布式共享存储架构，通过高速网路连接数据库计算节点和存储节点，引擎特别为高速网络以及共享存储架构重新设计，使POLARDB实现了即时生效的可扩展能力和运维能力。
我们来看一组数据，100%兼容MySQL 5.6、100T级别容量无缝扩容、2分钟实现只读副本扩容、1分钟实现全量备份、性能提升6倍，读性能达到100万QPS，写性能达到13万TPS。
要说为什么POLARDB一定要实现这些性能顶点？阿里云数据库掌门人余峰说，今天阿里云做的是下一代数据库，出发点就是希望未来几年用户都触及不到这个天花板，留下时间让他们继续创新，不拿出点真本事，何谈引领一个时代。
POLARDB实现了节点间共享存储，让数据库可以实现真正的秒级水平扩展，以往这一操作往往需要2天甚至更长。软硬件一体化的优化，使POLARDB相对于传统数据库相同的更新操作减少了50%的磁盘写入量。并缩短了写数据的路径，写性能显著提高，读事务实现优化。
新的3DXpoint存储介质、NVMe SSD和RDMA网卡等最新的软硬件优化技术，通过自研的分布式共享存储引擎Polarstore，实现了低成本，高性能的突破。在标准场景下，POLARDB性能是MySQL的6倍，单实例最大存储容量实现100TB，而在云数据库厂商中，POLARDB性能超越了友商的“阿罗拉”1.2倍，实现了2倍于其单实例最大存储容量的超越，成为世界级的云数据库引导者。
云计算厂商必须要突破的技术壁垒
阿里巴巴，对于数据库的认知可谓异常深刻，早在淘宝初创时期，阿里巴巴就聚集了中国最顶尖的数据库团队，在那个远没有今天发达的时代，解决了一个个海量客户带来的问题，被国内数据库行业称为“黄埔军校”。
但要从根本上解决问题，则还是要回到数据库架构技术的突破上。
在经典的数据库架构中，一个是scale up，一个是scale out。
scale up的优点是能够迅速的通过硬件叠加的方法快速扩容，这是传统的商业数据库中经常采用的一种扩容手段。缺点则是随着容量的增大，成本将会成倍增长，对于许多初创企业来说，这样的成本很难接受。
scale out则是通过分布式技术来解决数据库容量的问题，这个思路来自于谷歌，优点是成本相较于scale up更低，然而缺点也显而易见，需要对整个应用的底层进行改写，对于许多复杂的系统来说，这无异于是一场灾难。
而POLARDB通过计算与存储分离的革命设计，既拥有分布式设计的低成本优势，又具有集中式的易用性，从底层解决了海量数据扩容的问题。同时，通过分布式存储技术，使数据库的成本远低于商用数据库。
正是有了这样低成本，高性能，高可靠性的数据库产品，未来中国才会诞生越来越多的爆发式独角兽，而无需担心高昂的数据库成本与扩容的问题，最大化地利用互联网的能力、IT基础设施上轻装上阵，全面云化，用更低的成本、获得了更好的IT处理能力，这才是未来企业的数字化答案。
就在不久之前，据媒体报道，众安保险将以750亿元的价格在香港上市，而很少有人知道的是，众安保险从诞生的第一天起，数据就在阿里云云数据库上。而在全国的各个城市，如火如荼的共享单车背后，云数据库支撑着oFo在三个月中从数万用户增长到千万量级。
In today's cloud database market, such stories are constantly happening. Behind them, Alibaba Cloud brought POLARDB with its practical experience of more than 100,000 database instances.
Original link: https://yq.aliyun.com/articles/223456