1. primers

Someone left the following comment on an SQL-focused public account:

Reason 2.

In the last decade, the field of data management has changed dramatically. Hadoop, Spark, MemcacheDB, Redis, Vertica, HANA, Cassandra, MongoDB, InfluxDB, TiDB, etc. All kinds of high-end summit, forum really only AI, big data, it is really rare traditional relational database figure.

There are four reasons for this preliminary analysis:

  • one

The rise of cloud computing and the widespread use of open source databases have dramatically reduced the cost of computing power and data storage. Take Oracle, the leader of traditional relational data, as an example. With the rise of the Internet, massive users produce massive data. However, when Oracle is deployed on a large scale, not only Oracle software itself is expensive, but also the storage cost of IBM minicomp +EMC with it is huge. For example, Oracle’s data cluster is moved from one room to another, charging tens of thousands of yuan. Hardware and software maintenance costs, including requirements for administrators, are also very high.

Oracle also has a performance ceiling, followed by a closed-source black box. No matter how hard you try, you can’t predict the problems that might arise in a scene you haven’t encountered before.” Most of China’s Internet companies are from a rough background and attach great importance to cost performance. Yahoo, the Internet leader in the United States, took the lead in using MySQL database, and Chinese Internet companies followed suit.

The prime example is Alibaba. Alibaba was once called Oracle’s “star customer,” the “Huangpu Military Academy.” However, as the number of registered users increases, more and more data are generated by users. Taobao has started the largest Oracle RAC cluster in Asia. Due to the large amount of data and business requirements, the DATA volume of Alibaba B2B Chinese website maintains a 98% CPU utilization rate from 08:00 to 09:30 every year, and the server load is also extremely high. Even if hot and cold isolation is adopted, the problem of large data volume and large concurrency cannot be solved. The problem will occur again soon after replacement of storage devices.

Alibaba was forced to “save” a technical route to IOE: “Low-cost, linear controllable, decentralized (distributed) : IBM, PC Sever to replace the minicomputer; Go to Oracle, use MySQL instead; Go EMC, use mid – and low-end storage. In the subsequent explosive growth, including the business peak represented by the Double 11 in 2014, 15, and 16, the MySQL branch developed by Ali based on the open source version well supported the extremely strict business requirements. Finally, MySQL, which was very convenient to cloud and open source, gradually began to replace Oracle in China’s Internet companies. Become the first choice of many Chinese Internet enterprises.

– the second

It’s a reduction in the cost of common hardware. With the effect of Moore’s Law, the cost performance has been greatly improved over time. There are even HANA, Vertica, InfluxDB and many other in-memory or distributed in-memory databases. So just how cheap is cloud computing and cloud storage? Take a look at this AD I just pulled from a website today:

This kind of mass-marketed product can achieve real cabbage price due to the use of cheap PC Server, which was unimaginable 10 years ago.

  • The third

In addition, MySQL code and documentation are public, so the requirements of database development and maintenance personnel are greatly reduced.

  • The last

From SQL to NoSQL, databases supporting semi-structured data and unstructured data shine. The problem of unstructured data that could not be handled before has been solved, and unstructured data processing has become a standard configuration, which seems to occupy a hot spot in the market.

3. The status quo

So are traditional relational databases a thing of the past? If you think so, you are totally wrong! Let’s take a look at the authoritative data: we can see that from 2013 to 2019, the top three are basically Oracle, MySQL and Microsoft SQL Server, and they lead by a wide margin all year round.

This shows that while unstructured data is hot, relational databases are still sitting tight!

Why? Databases evolve in a colorful way, but ultimately the key lies in the business scenarios they excel at. Different types of databases adapt to different service scenarios. There are a lot of structured data in the real world, and the relational database is relatively mature to deal with them, and has advantages in all aspects. The focus of big data now lies in its effective solution to the problem of semi-structured and unstructured data, which has been unable to deal with before. Giving an unstructured database the role of a relational database is like sending a truck to a car rally.

In the past, relational databases were absolutely mainstream. Only in recent years with the development of 3G, 4G, pictures, audio and video content slightly increased. Later, with the development of 5G, AR\VR, Internet of Things and other new technologies, storage will become expensive, block chain and quantum computing will also bring new changes to the world…

Development of 4.

  • From single machine, cluster to completely distributed development, Oracle and other databases have been implemented
  • OLTP and OLAP fusion, this feature has been implemented for a long time
  • The integration of structured and unstructured, Oracle, MS SQL Server and so on has been implemented
  • Disk and memory database fusion, Oracle, MS SQL Server and so on have been implemented
  • The author’s preliminary study found that the fusion of different types of database, such as row, column, graph, object, document, content, timing, search engine, navigation class, has been partially realized
  • At first glance, data modeling tool support for the above features is also on the way, but far from perfect support

To sum up, in the future, the development of vertical relational database is relatively mature, at present, it is mainly in the horizontal development; Non – relational database mainly in the vertical further do deep – penetration.

5. Conclusion No matter how the world develops, I think relational database will remain the mainstream in the next ten years and will play an important role in the far future. Because structured data will remain humanity’s most important digital asset for the foreseeable future, relational databases will take over.

So the bottom line is: it’s not that relational databases aren’t important, it’s that there’s not much new to say; Unstructured data still has a lot of room for development, so people pay attention to it.

Light see what lively learn what, what fire learn what, impetuous state of mind can let a person technology learn deeply very hard. Practitioners need to figure out what’s going on in the world and what’s going to happen. The other key is to choose the right direction, to sit down, to get in.

Follow me for a weekly article about data related dry stuff.