A “graphic theory” is like the sea.

Edit | Shi Yaqiong

According to Gartner’s forecast, the global non-relational database (NoSQL) is expected to maintain a high growth rate of around 30% from 2020 to 2022, much higher than the overall database market. As a kind of non-relational database, graph database has obvious advantages in data storage and association, and it is a technical trend to deal with complex data relationships in the future. Fama Technology, which is introduced today, is a company that specializes in graph data technology and provides high-performance graph data storage and analysis platform for enterprises. It has two core products independently researched and developed – LightGraph and graphs. Computing platform (PandaGraph).

First, let’s take a brief look at the history of the development of the map.

Graphic theory originated from the study of the seven bridges of Königsberg in the 18th century, and helped many people solve many practical problems through the development of many mathematicians and even computer scientists. Today, graph-based data analysis methods have been applied to many scenarios of the Internet, such as social network analysis, web page sorting, community discovery, and are widely used in natural sciences such as computational chemistry, astrophysics, and bioinformatics.

The graph database does not refer to the database that stores the images, but rather stores and queries the data in the form of graphs. Nodes and relationships are the two elements that make up the graph. The graph calculation is to model the data according to the graph, to associate various types of data, and to fuse different sources and different types of data into the same graph for analysis. Get results that were difficult to find in the original independent analysis.

The emergence of graph databases is the result of information explosion and data complexity in the Internet age. Because the traditional relational database represented by Oracle can only represent the relationship in the form of two-dimensional tables, it is less efficient in dealing with complex relationships, and the graph data compared with it is better at dealing with complex relationships. In addition, the Fermat team has accumulated more than ten years of technical experience in graph data technology and has mastered a large number of research results. Therefore, the team hopes to put the technology into application to solve the real problem, which is also the entry of Fermat into the graph database market. the reason.

Starting with the data platform,

Ferma Chart Data Platform

In terms of product performance, Fermat has a technical advantage compared to its competitors:

  • Graphic Data Platform (LightGraph): The current speed can reach 10 million vertices per second, enabling 10TB external storage. Has the ability to store, query, and analyze easily. LightGraph performs much faster on simple tasks than Neo4j and TigerGraph.

  • PanaGraph: The advantage of the Fermat diagram analysis platform is its speed and memory. In terms of processing speed, PandaGraph is 300 times faster than Spark and GraphX; in memory usage, PandaGraph saves an order of magnitude (12 times) over Graph X.

In addition, Fermat’s competitive advantage in the graph calculation and graph data market is reflected in both team strength and market space:

Company said: Fermat has invested a lot of research and development in the field of graph calculation and graph data. At present, the company has a total of 16 employees, 13 of whom are R&D. The core team comes from Tsinghua University and well-known communication equipment manufacturers. The company’s annual research and development costs account for more than 80% of the total cost.

According to the Gartner research report, the graph database market will grow at a rate of 100% per year from 2019 to 2022; in the past 2018, the graph database market is about $500 million, and conservatively estimated at least 80 by 2022. One hundred million U.S. dollars. The market is still in the blue ocean.

In the profit model, Fermat offers standard software products with relatively common technology, so there is no need to customize the product for the industry. The procurement cycle is about one year, mainly relying on bidding.

The company currently serves 7-8 large and medium-sized companies with an average customer price of approximately one million. About 80%-90% of them come from the financial sector. The company has recently developed customers in the industrial sector and will develop small and medium-sized customers in the future. In the early days, the company used direct sales as the main method of obtaining customers, and is currently developing channels. The company expects to generate revenues of several million this year and reach 20 million next year. The main growth point comes from the map database, and has been talking with a number of large banks.

Although graph calculations and graph data have performance advantages and broad market space, they still face difficulties in technology and business models.

For graph data, graph data needs to be applied on a large scale, and the difficulty is first and foremost in technology. Because the storage and analysis of graph data is different from the traditional relational database, the data regularity is high. Once the data is poorly organized, it will affect the calculation, and the analysis difficulty will increase with the increase of the analysis dimension.

In business, for graph computing, graph computing requires high data comprehensiveness, because multiple data sources are involved, and data sources tend to belong to multiple systems, so data integration is a problem, which is also All the problems facing big data platforms. Although the graph database product can technically guarantee the ease of use of the tool, in general, it is necessary to coordinate the departments in charge of different systems to cooperate with the work, which requires a great labor cost.

In addition, the current market calculation market has not been opened, Because the charging model is a difficult point, mainly because it is difficult to price the product. The frequency and importance of customers are not directly proportional. For example, sorting calculations on web pages involves trillions of web page data processing. It is too expensive for customers to collect money according to the amount of data. Moreover, the computing system may only calculate one or two times a day, which is difficult to count according to the number of machines. Money, and the idle computer can also cause high operating costs.

Graphic data is facing global competition, and current competitors include:

TigerGraph in the United States (established in 2012, current A round); domestic companies including BAT, these Internet head companies will develop corresponding map databases for their own products, but the application of this graph database is more targeted So it does not apply to other customers on the market.

In May 2016, Fermat Technology received investment from investors such as Fengrui, Qingdao Taihao, Jinshajiang and Jinyun Fund, with a post-investment valuation of 80 million. In November 2017, it acquired Jingdong Financial’s pre-A round of investment. The post-investment valuation is about 160 million.

The company said that graph computing will expand from the financial sector to the industrial, manufacturing, and Internet sectors in the application field; graph data currently lacks standard specifications in the industry, and the company is actively participating in industry standard customization, hoping to become like a traditional relationship. A standard database like the database. At present, Fermat’s graph database can support tens of billions of nodes, which is suitable for most customers in the market. However, if large-scale data is involved in the future, distributed databases may be needed, and the company is currently in the research and development stage.