View metadata management by two instances

View metadata management by two instances

 

Designing a Bi system inevitably involves dealing with metadata, but sometimes people cannot feel it. For example, to understand the database structure, you do not need to view it from the system table. You may only need a document, command, or database administration tool to view it; however, if you want to develop a queryer for use by semi-business and semi-technical personnel to drag and drop them into SQL queries based on business terms, such a queryer will inevitably access metadata.
What is it?
What exactly is metadata? The definition of "data about data" is indeed very simple, but it does not seem accurate enough as an iterative definition. In terms of name, metadata is also data, so what is the data that describes metadata? Metadata? This iteration is endless.
This definition may be clearer by using an analogy. Think of metadata governance as a customer governance system. In order to better serve customers (in fact, how to earn more profits from customers), enterprises need to manage customers and improve customer relationships. In the same way, the metadata management system also aims to make better use of data. The customer has a life cycle, such as when the customer is served by the enterprise, when the customer is detached from the enterprise, and what status the customer is in. The same is true for the data, when the data is generated, and when it is used by the user, status changes.
In a data warehouse, the concept of metadata has been enhanced. in the overall architecture of each data warehouse project, almost all of them have the "metadata governance" module to traverse other modules. Obviously, this indicates that it is a basic module that can serve other modules such as OLAP and ETL. But in fact, it is rare that a completed data warehouse project contains an independent metadata section. Most projects and metadata are scattered in various Bi tools.
These scattered metadata are inconsistent. For example, the structure definition of a table may appear in ER design tools, and of course in the database data dictionary, it is also possible to define the source and target of the ETL tool. So many repeated definitions, of course, will lead to data inconsistency, but also leave a broad space for metadata governance tools. Their role is to centrally manage these scattered metadata, like a data warehouse, data is collected from different sources, including ETL, cleaning, and even re-modeling.

Who did it?
For some large enterprises, although sometimes the role of establishing a metadata governance system cannot be determined, there are still some requirements in this regard. For example, China Mobile has two provincial companies in Jilin and Hubei provinces to publicize their metadata governance projects in a high profile. This is an active attempt and is also promoting the application of metadata governance across the industry. From the publicity text, it seems like a high-sounding and boring, but another D-province company has a comparison report with these pioneers, but it is quite interesting.
Provincial D companies also have metadata governance content, but they have not yet formed a system, this report compares the organizational structure, construction time, project investment, and system functions with Jilin mobile metadata system. The architecture and investment are relatively simple: the metadata management system of Jilin mobile was built later, but it was built as a project, in contrast, Province D is built in a management analysis system. Therefore, the organizational structure must be different. In terms of project investment, Jilin mobile has invested millions, because the provincial Company D has not established a separate project, the report shows that there is no penny.
The most important difference lies in system functions. As a project, Jilin metadata provides an easy-to-use front-end. Company D focuses on the back-end metadata management and provides technical applications such as system monitoring, governance, and optimization, it is intended for technical personnel, with little consideration for business metadata and almost no interface provided for business personnel.
Regardless of the comparison results of the two provinces, the two companies still lack in-depth application of metadata governance, such as data quality governance, impact analysis, and lineage analysis, most of them are emphasizing the platform of metadata governance. The so-called platform is to set up the platform, and no matter what drama is sung above. Whether the platform can support singing or not is smooth, and there is nothing to measure.

How is it?
In general, metadata governance is still an immature field. The specific reasons are business and technology.
From the business perspective, many people are not clear about the purpose of establishing a metadata governance and exchange platform, and no one knows how much value these metadata brings to the Enterprise.
From the current stage of domestic enterprises, the establishment of such a platform does not produce much value. There are a variety of arguments about the platform's initial intention and who will use it. For example, some people say that this is the premise of enterprise data standards, and some people think that it is the basis of enterprise data integration, which can make the system scalable. It sounds reasonable, but it's too empty. If you keep asking questions, these statements cannot be elaborated. For example, what is the purpose of establishing enterprise data standards? What is the purpose of data integration?
This questioning does not mean denying data standards or data integration, but means that its purpose is not direct enough. You can use metadata governance as an enterprise's infrastructure, just like city road building and afforestation, to achieve a ambitious goal and benefit the masses. However, for general enterprises, when the entire IT system is not mature enough, it is undoubtedly too advanced to talk about the infrastructure. There is no problem with metadata governance. The key point is who will you use? What are functions? If someone asks for data quality control, performance optimization, system monitoring, flexible REPORT query, and so on, these can be done directly, but if you stand on the cloud and say "I want to establish enterprise data standards", you should first open up metadata and look at how to solve the communication gaps and work habits of different departments.
Technically, unified metadata standards have not yet been truly established.
Generally, metadata governance tools mostly provide metadata exchange functions, based on a certain standard, from other bi tools (such as er modeling tools, data warehouses, ETL tools, OLAP tools, etc) the metadata is extracted to the centralized database. Currently, the standard of the existing metadata is CWM (common warehouse model ).
But the problem is that this type of tool involves interaction with many tools. Even if it complies with a certain standard, the problem still cannot be solved if other tools do not comply. For example, the metacenter of DAG (Data advantage group) is a leader in the metadata governance tool field, but its functions are also very limited. It is applied in the project team. It can indeed extract metadata from products such as Oracle and corner stone. For ER modeling tools, it can extract table design information from Erwin and has embedded extraction modules. On the other hand, there is no extraction module for powerdesigner. The problem is that the project uses powerdesigner to design the ER model. Therefore, if such tools cannot completely concentrate all the metadata, data streams in the entire data warehouse may be faulty, so consistency analysis and lineage analysis cannot be established.
Currently, the concept of data as an enterprise's information assets has been gradually accepted and valued. The desire of business personnel to understand data and the requirement of the Administrator on data quality will make the metadata more practical and important.

Source: chinabi

Link to this article: View metadata management by two instances
Reprinted please indicate the source: Business Intelligence blog-dinosboy