– Introduction This article is all guesswork, do not take it seriously.

Two months ago, there was a news that OceanBase, a database owned by Alibaba, broke the nine-year record and won the championship. Its performance was twice that of Oracle, the second largest database in the world. Therefore, “China’s homegrown database broke the western monopoly”, science and Technology Daily commented: “The ranking of the international authority is also a sign of China’s database changing lanes to overtake in the cloud era.” But then it was said that Oracle “can’t or won’t.” Alibaba won, but the meaning was limited, but the details were vague. As a database veteran of more than 10 years, I try to answer the following questions for you;

  • What exactly is tPC-C, the database’s world record?
  • Did OceanBase beat Oracle or not?
  • Some say Ali’s record is a pile of hardware. Is that true?
  • How significant is Alibaba’s rise to the top?

1. What exactly is TPC-C? To talk about TPC-C, you need to know TPC first. Transaction Processing Performance Council (TPC) what is TPC? It’s actually the international mobile phone review organization, sorry, computer systems review organization, it’s a non-profit industry organization. It can be simply understood that its responsibility is to publish the enterprise database evaluation list and develop the test standards, procedures and specifications for the list. Database products from oracle, IBM, Microsoft and others have participated in this review several times. It was founded in 1988 and is headquartered in the United States. TPC currently (as of January 1, 2020) has a total of 13 categories, focusing on performance and value for money. TPC list examines the most real, the most core things, greatly put an end to cheating, very scientific and very clever software and hardware architecture do not limit, independent and objective evaluation there is a 60-day publicity period, in addition, there is absolutely no economic interests affect the list results. Because of its international, professional, strict and objective, after years of testing, it has become the most authoritative list in the industry. Simply speaking, it is equivalent to the “World Cup” in the database field.

Tpc-c is the most important standard published by TPC, because it mainly measures the performance of online transactional applications. Colloquially speaking, the so-called ONLINE transaction database (OLTP) refers to the common but very important core business systems, such as 12306’s train ticket selling system, Meituan takeout ordering system, etc. Corresponding to the OLTP system is the OLAP system, which can be commonly known as the report system. For example, the report system of China Mobile helps users to count the phone charges or the comprehensive profit of each month. Now, of course, big data-related standards have been added. Generally speaking, it doesn’t matter if the statements come out earlier or later; But I placed an order, half a day can not complete the transaction will affect the business to make money. In addition, OLTP is also the basis of OLAP, so the industry pays more attention to online transactional databases, so TPC-C is relatively the most important. This time, Ali won the tPC-C list. In 2006, The State Council issued the Outline of The National Medium – and Long-term Scientific and Technological Development Plan (2006-2020), which first proposed the concept of “core and high base”, in which “base” refers to the basic software products with the core such as database, operating system and middleware. Customers in key sectors such as finance, telecommunications, and government generally refer to TPC-C results to measure transaction capabilities of various database vendors.

So how professional and rigorous is TPC-C? Just two examples:

  • Tpc-c simulates real-world business systems, which you can think of as a simplified version of luckin’s mobile App for selling coffee. Tpc-c has five types of transactions: order creation, order payment, order query, order shipment, and inventory query. Each transaction has a specified proportion, in which order payment should not be less than 43%, order inquiry, order delivery and inventory inquiry should not be less than 4%, and the rest is order creation (not more than 45%). 1% of the order creation is to be rolled back. Maximum 15 items, average 10 items. The model is based on a warehouse with 10 points of sale/warehouse each serving 3000 customers. If a customer goes to the point of sale to buy something, it may be 3 items, it may be 12 items, because not all the items are in the local warehouse, so each item is assumed to have a 1% probability of being in another warehouse, and each transaction created by an order in a distributed system has a 10% probability of being a distributed transaction. There is also a 15% probability that the order payment will not be paid in the local warehouse, but will be paid in the remote warehouse, which has also become a distributed transaction. The tpmC value is the number of order creation transactions executed per minute. Finally, the main assessment tpmC with cost-effective.

  • The entire application and database should be open, what machine you’re using, what the configuration of the machine is, and how much it costs. Bar essence, no, challenger is available on the open market. From functions, data consistency and disaster recovery capabilities to the price of the mouse and keyboard used in the testing process, strict disclosure is required.

2. Did OceanBase beat Oracle? You can call it defeated or you can call it not.

First of all, why did OceanBase beat it? It’s easy. OceanBase has topped the list through tPC-C, an open and rigorous testing standard. You Oracle bull then you brush back on your ability. Now THAT I’m the champion, I’m the best! This is called openness, fairness and justice. So Academician of Chinese Academy of Engineering, famous national treasure level computer expert Li Guojie expresses to this, this is the great breakthrough that China basic software obtains!

Looking at the price of a single tmpC, OceanBase’s price is 6.25 YUAN, and Oracle’s is 1.01 dollars, nearly 1 to 1. However, OceanBase’s hardware cost only accounts for 18%, while Oracle’s costs are 65%. So OceanBase is also very good in terms of value for money.

So why not? This is separated from performance, scenarios, accessibility, and ecology.

– Performance champion Oracle hasn’t updated this list in 9 years, and hardware has changed a lot in those 9 years. The server side is not exactly Moore’s Law, but it should be several times better for the same cost.

Another question many people ask is, why hasn’t Oracle refreshed the list in so many years? In the past few decades, everyone was playing with it every year. If you exceeded 20%, it was very scary. IBM DB2 cluster only managed to grow to more than 10 million, but Oracle added to more than 30 million in 2010.

What’s more, Oracle made 5.06 million and 8.55 million PCS in 2012 and 2013 respectively. Oracle has used 27 machines to do more than 30 million cluster (is the ali exceeded this), why do the following to do 5 million, more than 8 million single results? It is estimated that the original achievement of more than 30 million tmpC will be more than 1 million tmpC. Now, I can achieve 4 to 8 times of the original achievement on a single machine, and finally only achieve 5 to 6 times of the original achievement, or nearly 180 million tmpC, which of you can do better than me? Other manufacturers a look, I le go, make this thing cost money and cost old strength, you make me so how to play?

One after another, Oracle is too lazy to brush the top – anyway, I have the world’s first, you who dare to fight with me? Ha ha ha!

– Scenario Oracle provides good support for OLAP and unstructured data integration. OceanBase still has a lot of work to do in these important areas. All of these scenarios are very important today, not to mention mobile, real-time computing, in-memory computing, etc. Other specific functions, auxiliary tools and hardware and software from other manufacturers, not to mention.

Some people say that Ali’s record is a pile of hardware. Is it really so? You don’t know anything about TPC.

On the one hand OceanBase is distributed, so of course you heap hardware.

On the other hand, if pure hardware can be piled, won’t the review end up being a game about which vendor has the money? TPC is mainly a technical evaluation, in order to prevent manufacturers from pure pile hardware, there is a special test index on this aspect, that is, the total cost of system hardware and software, including technical support services, allocated to the price of each tpmC in 3 years. This index OceanBase 6.25 RMB to Oracle 1.01 USD, considering the exchange rate factor, basically close to 1 to 1; But OceanBase’s hardware cost is 18%, while Oracle’s is 65%. So hardware cost is actually where OceanBase has the upper hand. Data reliability is certainly not a problem.

4. How significant is Alibaba’s rise to the top? OceanBase’s victory over Oracle, the database leader for nine years, is significant.

  • brand

By supporting Double 11, OceanBase established its brand among techies; However, in the commercial market, the success of this ranking is undoubtedly a huge brand bonus to the database, which should be the main purpose of Ali this ranking. In addition, this is also a big plus for the Alibaba brand.

The tPC-C professional and rigorous test certifies the basic ACID capability of the database under OceanBase’s innovative distributed architecture, and greatly strengthens the recognition of the database itself by all parties.

Of course, Ali itself through this action to further improve the product, but also exercise the team, to eliminate internal doubts.

This is the first time a Chinese company has topped the list. Oracle is 42 years of accumulation of products, Alibaba only took 9 years to do so, under the current situation, also inspired the Chinese people.

  • Technology innovation

Distributed technology breaks the performance bottleneck of single machine and cluster, and verifies the technical feasibility of distributed relational database through strict test.

Also because of distributed technology, it is possible to use ordinary machines to store critical data. Because it is not a special device but a cloud host, it can be easily rented, which greatly reduces storage costs and test costs. Tpc-c requires the overall cost of software and hardware plus technical support for the whole system for 3 years. OceanBase only rented Ali Cloud for 3 months this time, and the total test hardware cost is 380 million ×18%÷36(converted into months for 3 years)×3= 5.7 million RMB. Oracle’s test hardware cost 30,249,688×6 (exchange rate) *65%= 118 million, approximately 20.7 times. The hardware Oracle tests can of course be used in other ways later, but you’ll have to pay for it before you test it.

  • Perfect the TPC

This is the first time in the world that a distributed database participates in TPC-C evaluation, and TPC takes this opportunity to clarify and improve the evaluation standards and methods of TPC-C.

Alibaba made a way for domestic database to hit the list, opened source the test tools, relevant documents and data, and shared the evaluation platform, which greatly reduced the cost of software and hardware for each manufacturer to participate in TPC-C evaluation, and greatly reduced the threshold.

  • Low dependence on foreign technology

OceanBase is a distributed relational database owned by Alibaba and Ant Financial with 100% independent intellectual property rights. OceanBase does not rely on any open-source database or relevant code of storage engine. Instead, it has developed distributed database product technology independently from the first line of code, with low external dependence.

Follow me weekly to share data modeling related dry stuff.