TDengine’s distributed clustering function has been open source, and it has been ranked number one on GitHub’s global Trends list for several consecutive days. People are wondering how the company can survive when core features are open source. Now I will share a blog I wrote in the company in October last year, hoping that the open source strategy of Taos data can give some inspiration to China’s 2B software industry and the open source community.

1: Purpose of open source

As a big data platform for the Internet of Things, TDengine belongs to basic software and can be classified as a database and big data tool. For users, the migration cost of basic software, especially database software, is extremely high. Therefore, it is very difficult to promote new basic software even if its performance and functions are super strong. In principle, the promotion and sales cost of a new product is more than 5 times that of a mature monopoly product, so the cost performance of a new product should be more than 5 times better than the existing product in the market before it can have an impact on it.

Open source is a good way to promote, and the most successful examples are Linux and MySQL. In the early days of open source, these two products were not comparable with professional products such as Sun Solaris and Oracle in terms of technology and products. However, as open source was free, it was welcomed by developers, and many developers also participated in it. Therefore, a certain user group was formed, and the user group gradually expanded, and eventually became the mainstream products.

On the other hand, the basic software must be geared to the global market, and it is difficult to make it big if it is limited to the Chinese market. The Chinese market accounts for less than 5% of global sales of basic software due to piracy, undeveloped awareness of payment and service. We cannot ignore the huge overseas market. And the basic software belongs to the typical technology products, the use is no national boundaries of the region. At the same time, for the operating system, database software, only the world’s top three can survive. Hence tao’s determination to target global markets from the day it was founded.

When it comes to overseas markets, marketing is more difficult, and the average developer will have a lot of doubts and distrust of basic software from China. Open source is the best way to increase trust and remove doubt. At the same time, open source can attract the attention of developers from all over the world, some of whom can become contributors, which is the best way to attract global talent for collaborative development. In the database world, the familiar open source software ClickHouse was developed by the Russians and Redis was developed by the Italians, but both have become popular worldwide.

With RedHat being acquired by IBM for $34 billion, MongoDB, ElasticSearch and others going public, Open source software, especially the Open Core model, has proven to be a successful business model. On the other hand, all the popular temporal databases in the market have open source community versions. If we do not have them, it will be more difficult to promote them. From this perspective, it is impossible not to open source.

In short, we open source to solve the problem of market promotion, the purpose is to quickly gain market share, or even market monopoly position, to form a brand.

2: Select a License

To be clear, open source does not mean code is free to use. Users must accept the provisions in the License, otherwise it is a violation. There are many licenses for open source products, the popular ones are Apache, BSD, MIT, GPL, etc. We chose AGPL, which is an enhanced version of the GPL. The GPL is a dual license. If you want to use the GPL for free, your code must be open source. Otherwise, you need to pay for the GPL. The AGPL comes in the context of cloud computing and is much more restrictive.

Our goal with AGPL is to prevent cloud vendors from offering services with open source code and monopolizing profits, rather than forcing other users to open source their software. Because cloud computing is the trend, the market share is getting higher and higher, and the market will only accommodate a few cloud vendors. If cloud vendors organize their own technical support teams, Tao data will lose an important way to make money. But we are happy to work with cloud vendors to provide TDengine services and share the profits. MongoDB, which uses the AGPL, started using its own open source protocol a year ago with more restrictions, in order not to allow cloud vendors to use it for free.

One more thing to make clear is that our open source protocol can be changed. When we have time, we will develop an open source protocol that only limits the use of cloud vendors, and only the release version will be implemented according to the open source protocol at that time. AGPL is the simplest and most appropriate protection method at this stage.

3: What needs to be open source?

There are two kinds of open source, one is 100% open source, and the other is partial open source. For 100% open source, RedHat is currently the only commercially successful company, while there are many commercially successful companies that are partially open source. The current trend is Open Core, which is to Open source the Core code.

We are a company that wants to be commercially successful, so we need to adopt a business model that everyone agrees on — Open Core. For TDengine, we need to open source the core modules with real technological breakthroughs, including: storage engine, query computing engine, RPC, and so on. If these core modules are not open source, it is difficult to convince developers to use them because they have many alternatives.

When the open source was announced in July 2019, we decided to make the cluster closed source from the perspective of subsequent publicity and commercial expansion (note: the cluster function was just open source on August 3, 2020). But from talking to Steven and a lot of other companies in the US, we should open source it. There are several reasons for this. First, the open source versions of Our competitors InfluxData, Promethus and TimeScale do not offer clustering, so if we do, our products will be more competitive. 2. The time series data processing market is still in the Warring States Period. The definition of products and the market pattern are still changing. If we want to stand out in this age of confusion, we must be one step ahead and dare to do what our rivals dare not do.

In principle, all basic functions, especially the modules that demonstrate our technical strengths, need to be open source.

4: What doesn’t need to be open source?

What doesn’t need to be open source? In principle, only large customers have the required functions. For large enterprise customers, there are many needs, there is no technical challenge, and it is not the core module. They can develop by themselves, but limited to the cost of manpower and time. They are willing to purchase. The TDengine has the following features:

  • Data encryption: Data files are encrypted and saved
  • Remote Dr: A virtual node group can span several equipment rooms
  • Audit: View the various operation records of the DBA
  • User authentication interface: Supports LDAP and Active Directory
  • Multi-level storage: saves storage costs
  • Multi-tenancy: This is typically only required by large enterprises or those providing cloud services
  • System monitoring: The monitoring of TDengine should be incorporated into the overall monitoring system of the enterprise to reduce the operation and maintenance pressure
  • Graphical background management interface

In principle, ancillary features do not need to be open source.

5: After open source, what to sell?

Open source sells accessibility: the basic core features actually work, and they work well. But any large-scale enterprises, as IT leaders, must consider the security of the data (including disaster, tolerance, audit, encryption, audit, etc.), considering the complexity of the operation and maintenance (monitoring system can be integrated with existing IT, whether to have professional management tools), these functions, they can be based on open source version of the development, But the cost of development and maintenance must be higher than buying the original products. Large Internet companies, such as Google, Alibaba, Tencent, etc., because of their huge data scale, they will be willing to organize teams to develop on the open source version, so it can be concluded that these big Internet companies will not pay to buy our products. But we still need to encourage them to adopt, because they are the model customer and can be used to spread the word.

Open source sells insurance and services: the CIO/CTO of a large corporation doesn’t save the company millions of dollars a year, leaving the entire data system in an unreliable state. Open source software is fine, but if there is no original support, if there is any technical problem, if the entire system is down for a few minutes, or even an hour, let alone a day, then the cost is huge, and even the HEAD of IT is fired. No one can guarantee that software will not break, this is the consensus of all IT people. Therefore, as long as a professional and experienced CIO/CTO, he will certainly buy professional products and services, at least when something goes wrong can find someone to blame. Therefore, during the sales process, we made it clear that as long as they are free customers, we can only provide services on GitHub or the technical community, and we cannot guarantee real-time service. But for paying customers, we have people who respond in real time. Even if the customer does not bump into a BUG in the process of use, only the original manufacturer knows how to do the optimal configuration and better integration with other systems, including technical training, software upgrade, etc., which is also the service we can provide.

6: sales strategy after open source

After open source, the general customer can use open source version. But some customers who need to use our closed-source capabilities, or just want our technical support, will need to buy our enterprise edition. The pricing strategy for enterprise edition is the same as traditional software, but we need to emphasize the value of the service and try to sell on an annual fee basis.

But open source has changed the way we sell, and the average customer has become a free customer, leaving us with no profit. But these free customers form a huge user group of TDengine, they bring TDengine spread, provide rapid market feedback to TDengine, and intangibly reduce the market space of our competitors.

Through open source, our customer acquisition channels mainly come from online channels, and our customers who take the initiative to consult us have a certain understanding of the product, the PoC links will be greatly shortened, and the direct sales cost will be greatly reduced. Judging from the effect after we open source in July, hundreds of customers have contacted us actively in three months. For traditional 2B software sales, that’s a pretty staggering number.

But the selling itself hasn’t changed, open source just brings a lot of effective leads, and we still need salespeople to communicate with customers carefully. This kind of communication, before our brand position is fully established, in addition to telephone and wechat communication, also need to face to face, detailed understanding of their pain points, help them solve, there is no shortcut.

Therefore, our current sales strategy is: online drainage, offline transaction.

TaoJianHui

In Wangjing, Beijing, October 13, 2019