DAAS architecture and Implementation (I)

1. preface

DaaS Data as a service is a service pattern , I.e. data in the form of service , Provide value to customers , Participate in the customer's business , It is also a subdivision of software as a service . meanwhile DaaS It has the general characteristics of cloud computing , Including renting instead of buying , Pay as you go 、 Pay as you go .

In this paper, DaaS The architecture and implementation choice of , For enterprises with a large number of high-quality data resources , You can refer to building a data business line , And then realize the capitalization of data 、 Value . It should be noted that the various legends in this article are only logical diagrams , Are simplified .

2. Infrastructure

The system consists of four parts , Include ：

• The customer is based on API Self developed applications to access data .
• API Data interface , Encapsulates and abstracts the data definition and permitted data access patterns .
• Data service is API The concrete realization of the function .
• The database stores the original data , Of course, there is unstructured data , Such as some pictures 、 video 、 Proprietary documents, etc .

This architecture is suitable for two situations , One is a trusted environment , The second is the early commercial verification stage , The advantage is simple structure , Low cost of implementation .

3. + Official applications

In many cases , The customer has no application development ability , A lightweight graphical tool needs to be superimposed on the interface , Such as web pages 、 Applet .

4. + Application certification

increase App After authentication , Can alleviate the API Interface security anxiety . With the help of offline 、 Online authorization , License specific applications to access specific interfaces .

5. + Model

Based on the understanding of application scenarios , Process the original data to a certain extent , Generate model data （ Indirect data ）, Provide model data externally . This will simplify the customer app Development of , At the same time, it also protects the original data to a certain extent .

6. + desensitization

Some data cannot be directly disclosed , Some desensitization is required , Dynamic desensitization or static desensitization can be used . Dynamic desensitization is the calculation of access time , Static desensitization is completed in advance .

7. + In Library calculation

Traditional databases provide some computing power , For example, common statistical functions 、 Stored procedure, etc . Now the computing power of the new database is getting stronger and stronger , Instead of sending data around , Limited by bandwidth , It's better to drop the calculation into the database . meanwhile , This also reduces the issue data , Reduces data security risks .

8. + charging

There are many ways to charge , Here are two kinds of , One is in api The layer performs billing services synchronously or asynchronously , The other is based on offline statistical log data , Generate expenses . The former is more timely , But it requires high engineering capacity , At the same time, it will also reduce the efficiency of interface access ; The latter system is more robust , But there is a lag problem , This can also be solved through commercial measures .

9. + Development

When considering development scenarios , Will increase other needs , For example, the test environment 、 Production environment , such as SDK、 Developing documents . But the most challenging technology is the interface version .

reflection

• How to do Access Acceleration ？
• How to do high availability ？
• How to do elasticity ？
• Privacy computing 、 Federal learning 、 Multiparty secure computing 、 Trusted computing 、 What is confidential Computing ？
• After introducing these ,DaaS How will the architecture evolve ？

author ： Huangjunlei

Data driven value Love to achieve extraordinary