DAAS data as a service is a service model, that is, data provides value to customers in the form of service and participates in customers’ business. It is also a subdivision of software as a service. At the same time, DAAS has the general characteristics of cloud computing, including rent instead of purchase, pay as you need and pay as you use.
This paper introduces the architecture and implementation options of DAAS. For enterprises with a large number of high-quality data resources, we can refer to building a data business line to realize the asset and value of data. It should be noted that the various legends in this paper are only logical diagrams and have been simplified.
The system consists of four parts, including:
- Customers develop their own applications based on API, so as to achieve the purpose of accessing data.
- API is a data interface that encapsulates and abstracts the definition of data and the permitted data access mode.
- Data service is the concrete realization of API function.
- The database stores the original data and, of course, unstructured data, such as some pictures, videos, proprietary files and so on.
This architecture is suitable for two situations: one is in a trusted environment, and the other is in the early commercial verification stage. Its advantages are simple structure and low implementation cost.
3. + official application
In many cases, customers do not have the ability of application development and need to overlay a lightweight graphical tool on the interface, such as web pages and applets.
+ application certification
After adding app authentication, it can alleviate the anxiety about API interface security. With the help of offline and online authorization, specific applications are allowed to access specific interfaces.
5. + model
Based on the understanding of the application scenario, the original data is processed to generate model data (indirect data) and provide model data externally. This will simplify the development of customer apps and protect the original data to a certain extent.
6. + desensitization
Some data cannot be directly external and need some desensitization processing. Dynamic desensitization or static desensitization can be used. Dynamic desensitization is to calculate when accessing, while static desensitization is completed in advance.
7. + calculation in warehouse
Traditional databases provide some computing power, such as common statistical functions, stored procedures and so on. Now the computing power of the new database is becoming stronger and stronger. Instead of transmitting data around and limited by bandwidth, it is better to lower the calculation to the database. At the same time, this also reduces the outbound data and reduces the data security risk.
8. + billing
There are many ways of billing. Here are two ways. One is to conduct billing business synchronously or asynchronously in the API layer, and the other is to generate fees based on offline statistical log data. The former is more timely, but requires high engineering capacity, which will also reduce the efficiency of interface access; The latter system is more robust, but there is a lag problem, which can also be solved by commercial measures.
9. + Development
When considering the development scenario, other requirements will be added, such as test environment, production environment, such as SDK and development documents. But the most challenging technology is the version of the interface.
Alibaba cloud API gateway is a very good learning object, which can be used as a reference for friends who want to implement a DAAS system. Of course, instead of building a wheel repeatedly, it’s better to enjoy the products formed at present and focus on the development of core business.
● how to speed up access?
● how to make high availability?
● how to be flexible?
● what are private computing, federated learning, multi-party secure computing, trusted computing, and confidential computing?
● how will DAAS architecture evolve after introducing these?