Introduction: This sharing introduces how Kersuch cloud uses Ali Cloud big data products to build data center. Gueruyun is a comprehensive SaaS company established in 2012, covering catering, retail, beauty industry, as well as other formats and services. By 2020, Guruyun has served 600,000 merchants and helped 600,000 merchants realize digital and intelligent transformation. In the future, we will cover more merchants.

Li Hao, Technical Director of Keruyun

This sharing introduces how Kersuch cloud uses Ali Cloud big data products to build data center.

Gueruyun is a comprehensive SaaS company established in 2012, covering catering, retail, beauty industry, as well as other formats and services. By 2020, Guruyun has served 600,000 merchants and helped 600,000 merchants realize digital and intelligent transformation. In the future, we will cover more merchants.

At present, Keroyun is a four-center structure, with our R&D center in Chengdu, hardware R&D center in Shenzhen, headquarters in Beijing, and sales center in Wuhan. Next, we mainly introduce our business scope: Guruyun is based on software and hardware integrated SaaS cashier service as the core, so the hardware and software of SaaS cashier system is our first layer, which is our efficiency tool. On the second floor, we create an ecosystem of people, finance, goods and customers together with friends of catering and retail enterprises. At the third level, we use value-added services, such as marketing, supply chain, human efficiency, business intelligence, financial services, as well as big data applications, to meet the needs of businesses at all levels. From S1 to S5, we cover all brands of all sizes. Our vision is to help customers, help merchants to open the world, customers like clouds, and we can better serve merchants, help merchants to improve efficiency and reduce costs, to obtain more revenue, reduce more costs.

At present, the whole system of Guruyun is built on the service of Ali Cloud, which ensures that we become a platform-level company with less resources.

Next, we will introduce how to use Ali Cloud products to carry out data center construction. As we all know, Alibaba’s Data center is the core of three Ones, One ID, One Data and One Service. On the basis of the improvement of infrastructure, we first need to achieve data integration, so we use Ali Cloud DataHub, DataWorks, DTS and other products to unify our business data into our Hadoop cluster, and then we now migrate to our MaxCompute. Use big data products to carry out the construction of the whole data warehouse. MaxCompute helped us realize the entire offline data calculation and storage, including the construction of warehouse space. Then we used PAI to build our algorithm center and machine learning model, and Flink technology to build a real-time computing platform. On the basis of these real-time and off-line calculations, we have established a unified query service. The Hologres product of Ali Cloud is used to realize our One Service concept of integrated query.

On this basis, we ensured the realization of our entire data application, including our internal data application, BI products, external data application, large screen, reports, the whole algorithm, intelligent recommendation, precision marketing, etc., and established our data center like a cloud of customers. And then quickly meet the data needs of our internal and external users.

Next, let’s talk about how we use Aliyun’s products to help solve our pain points in this process.

The first thing we saw was that we had originally built our own Hadoop cluster, which was difficult to maintain because it was expensive. After one year’s effort, we changed the whole Hadoop cluster to MaxCompute. The effect is obvious, our operation and maintenance cost is reduced by 1 times, and the computing speed is increased by 8 times, which ensures our ability to deliver ETL and other computing processes quickly for use by the business line.

Second, we encountered a lot of data security problems, and the self-built Hadoop cluster could not do data audit, so we used the sensitive data protection product SDDP of Ali Cloud to classify data, protect our products and achieve zero data leakage. At the same time, our self-built Presto cluster cannot be integrated with our MaxCompute for interactive query. After investigation, we found that Hologres product has better performance than Presto, so we replaced Presto with Hologres product. This ensures that we can query directly from MaxCompute, making the whole interaction seamless. Then we met some data models and data modeling products. Currently, we are still investigating Dataphin products to realize the construction of the whole data model.

Our biggest pain point was the performance of the real-time large screen, and at this point we found that using Flink’s technology, coupled with QuickBI, could solve our front-end and back-end problems. Next we take real-time large screen as the core, to introduce how we apply, and then how to solve this problem.

The first difficulty we encountered was that we had too many data sources. We used MySQL, RDS, MongoDB, Redis, ES and other data sources. We needed to unify these data sources and solve the problem of too many data sources. The second aspect is the performance of our large screen. Now we have a very large amount of data, and a lot of styles and complicated requirements. How to solve this problem at this time? Then it is a big problem to see these requirements, generate these data, how to quickly display in the front end. So when we investigated the QuickBI product, we found that it did solve our problem.

Let’s see how we solved the problem. In the first question, we still mainly need to do data governance, source data management, blood relationship, and even some multi-data source processing to reduce our existing cluster. The second problem is that due to the large amount of data, all kinds of enterprises actually have a large amount of data, so they need to solve the problem of fast query. The solution is that we use the real-time computing platform of Aliyun and based on the open source technology of Flink to solve the whole problem of data query speed. I think the whole Ali Cloud Flink technology is indeed able to quickly query the massive data we want to query, its performance and high scalability we really get experience, I think Ali Cloud is still ranked the first in this field. In the whole big data front end, the problem we found was slow rendering. But we use QuickBI custom drag and drop to quickly define the data source, import into different boxes, can quickly query out.

The above is to solve some problems, the next introduction to the real-time computing platform is what kind of architecture. We set up our real-time computing platform through four layers: basic data layer, real-time computing layer, interface layer and display layer. Let’s focus on our real-time computing layer. In fact, we have some requirements for the computing layer, not only the data of the day, for example, the merchant needs to see how much our turnover is so far this month, not only the data of the day, but also all the data from August 1st to now. So it shows that we not only have access to flow tables, but also access to dimension tables, but also aggregation computing, so as to form a data flow. The convergence of multiple streams realizes that in the case of the invocation of the interface layer, the display layer can display the data that we can meet the needs of each dimension of the merchant on that day. Just like the example I gave, we can see the turnover of the current month so far, such a scene. So our real-time computing platform is mainly to meet the needs of all aspects of the business.

Let’s share a practical application scenario. We have already helped some vegetable markets to implement large data screens. This piece of display is mainly what the sales volume is that day, which sales volume is the best, so as to help merchants know what goods into the next day more appropriate. This big-screen application can actually help businesses get a good experience in real life scenes.

The real-time big screen of catering is based on our existing data, showing some of our Chinese catering big data, but this is only representative of some customers’ data. Of course, with more and more customers, we can do better and better, so that we can show what dishes Chinese people like, which dishes are the most delicious, and what kinds of tastes they like. We can find out through our real-time data calculation and data mining.

BI applications are mainly used by internal customers. Our operation team, sales team, R&D team and other teams can ensure that we play a significant role in data analysis, operation and decision-making assistance, saving a lot of costs and time for the management. With our BI products, we can solve such a need.

Summarize the main introduction of the three points, one is the customer such as cloud is what? Guest cloud is to help our catering, retail, beauty industry businesses, stores open the world, guest cloud to the vision of SaaS companies. The second point is how To use Aliyun’s big data products to build guerudun’s data center. The third point is how Keruyun uses real-time large screens, our business portrait products and other big data application products to empower businesses.

Data delivery experts tell you, how can the data architecture be layered properly?

2020-12-02

Looking at the alternative understanding of the data center from DataPhin

2020-10-21

A detailed explanation of Ali Cloud Data Center, an article to fully understand the big data “Internet celebrities”

2020-08-29