Aliyun released a run than AWS database, we chatted with the product owner


three years ago, Aurora AWS is announcing Amazon relational database (Amazon Aurora).

“at that time, Amazon Aurora we think are top of the world. “Director aliyun database Yu Feng sighed when 36 krypton in an interview. But at yesterday’s press conference, they release new products unceremoniously running over the Amazon on Aurora.

September 21 press conference, aliyun combed their entire family of cloud databases, including relational, non-relational database NoSQL to mainstream (for content caching and Web application scenarios, including key-value type, document type, and so on), data-oriented hybrid analysis database and AI, Across search and Internet scenes and sequence databases. For the convenience of the user database on the cloud, aliyun supporting migration management services and tools.

However, launch event, focused on, or mentioned above run over Amazon new products–aliyun Aurora since the Institute’s commercial relational database PolarDB. According to aliyun introduction, the database uses third-generation distributed shared memory architecture, achieving separation of compute nodes and storage node, the calculation engine and the storage engine are rapid scalability, a database to meet the mixing effect of multi-type database. At present, the product has started to open beta.

POLARDB third generation distributed shared memory architecture

database everywhere and soft one

“each company will need a database. “In the information technology world, databases are almost everywhere, small to a calculator, Bluetooth speaker, to a computer, with regards to communications or data storage and computing, are inseparable from the database. We can usually use a computer to understand that every command is used to the database.

so how do you measure the quality of database? Performance, reliability and security of the data, are the three most basic dimensions.

Yu Feng explained that database is the most important thing is to save things, and take it out when needed, save what is the efficiency, take and how (performance) was very important. For example, System Taobao, with millions of account name and password, while 300,000 people is normal, how to read at 300,000, will not slow, high-concurrency read-write database is a very important one.

he is a figure of speech, “your money in the banking system, might be one line of code, if this information is lost, are amazing. “It’s like the reliability of the data. As for the data, who can and who cannot be modified, this is security.

what are factors that affect database performance?

“database is a software integration things, involving many factors. “

on the one hand is the hardware. Data needs to be stored, relating to storage media such as disks; but not all in my hard disk, it will be very slow, so essential to memory; after out needs to be calculated, so the CPU is very high; over how to to the user involves networking.

on the other hand is the algorithm. Algorithm involves something more complex, Yu Feng did not detail, also cited a few examples.

first, the database is very important to support high concurrency, whom once more, easily affected, if isolation to ensure efficiency.

Furthermore was functioning and ability to roll back a transaction. For example, transfers as one transaction, in fact relates to three actions: first check a enough money, then deducted money from a, and finally added to b there, if any step is unable to continue, the transaction does not complete, and data to ensure that once an error has occurred, and can roll back to the initial state, there must be no intermediate state.

in addition, offsite disaster recovery case how to ensure transactional consistency, if synchronization in a three-node, if there is a transmission problem, how you want to handle.

above all, involves algorithms.

aliyun PolarBD run over AWS Aurora?

even if you don’t know what a database is, but by the database, you must think of the Giants in the field of Oracle (Oracle). The company even has become synonymous with the database.

as Yu Feng says, databases require hardware and software integration. Oracle’s drawback is that software is not only very expensive, if you want to maximize database performance, the need to maintain “+IBM mini +EMC Oracle database storage” this combination, subsequent costly, very good ability for administrators.

later, when the Internet giant Yahoo began to use the open source MySQL database, MySQL itself has a flexible and cost-effective advantages, together with the Yahoo model, MySQL database has become the choice of many Internet companies. With the development of cloud computing, cloud databases began to appear, which uses a low threshold, for quick extension. 2015 AWS Re: Invent the meeting Amazon announce Amazon Aurora. Aurora is a relational database that can be leveraged across 6 copies of the 3 available regional data, the design goal is to provide high performance and high availability (99.99%).

aliyun PolarDB launched yesterday, apparently is the subject of Amazon Aurora, which used the same design concept with the latter, underlying the same massive improvements made based on MySQL.

at yesterday’s press conference, aliyun and came running, expressed in the standard scenario, the PolarDB is MySQL 6 times, 100T single instance storage capacity and database vendors in the cloud, aliyun PolarDB implementing AWS Aurora performance of 1.2 times. Yu Feng explains, this is aliyun advantages, Polar uses the latest RDMA network card and details of the database has been optimized.

according to Yu Feng introduction, the entire PolarBD solution can be divided into three parts: compatible with MySQL database software, a new generation of computer nodes, aliyun third generation distributed shared memory scheme.

a new generation of computing nodes of CPU and network card, as well as storage-SSD disk with FPGA cards, is individually tailored for the database. These can be said is the embodiment of hardware in the database services.

in the algorithm section, aliyun highlights, is the third generation of PolarDB distributed shared storage solution.

Yu Feng introduction throughout the cloud vendor’s database services at home and abroad, in the storage section, after three generations of development.

the first generation of traditional databases running on a single machine, computing (CPU) and the storage bound, use this site, meaning that each compute node, to hang a disc, its problem is that each node storage space is limited.
the second generation central storage, compute and storage separation, centralized management storage, solves the problem of storage expansion. Because each calculation has a copy of the data, there are many copies of the data in this system, the cost is very high. In addition, the synchronization between different nodes of the data may occur, data consistency is difficult to solve. “EMC storage very much, but why do databases without fire, because delay is especially great. Amazon advantage is delayed, and Ali is very small. “

third-generation database is different from the second generation, is the very high efficiency of communication between compute nodes. This is equivalent to all of the compute nodes is a cluster, a node can infinitely increase, stored in a disk array. Aliyun introduction, product architecture, PolarDB uses shared memory between nodes architecture, database speed extensions, by redesigning the PolarStore file system to achieve the same amount of disk write data updates by 50% and shorten the path of write data, the write performance is significantly improved, To read transaction implementation was optimized.

database system described above, and just yesterday started public testing, how to use, also needs to be seen. However, Yu Feng, said general power less of the company suddenly changed database, unless the original database service is too expensive, data reliability and performance requirements of request or service runs well on new business in the new database, consider replacement.

in order to facilitate the customer database, Ali has unveiled a “price of IaaS, PaaS service” based cloud server, means that you buy a database that contains the cloud Server service, as with cloud servers only price. In order to encourage customers to use PolarDB, aliyun said aliyun MySOL database by a customer, transfer directly on PolarDB, charge even more cheaper.

this is the legend of subsidy? Well, this mode is drop … …

Go back to homepage.