Best practices

When using tags, it's important to understand current or potential company and use-case structure. Using this information, you can choose the right tags. For example, if you are building a data lake for the presales department, and you know that there are plans to expand the data lake to include data from the after-sales department, using the tag department will help identify the costs and performance of each department separately. Planning, cost allocation, and optimization of code or data can be identified with more accuracy. Without the department tag, if presales data needs 15 minutes for data modeling and aftersales data needs 45 minutes, developers must spend more time in root cause analysis. With the department tag, developers would know exactly where to look.

Tagging ontology

Business and technology together play an important role in identifying the right tags to use. From a business perspective, a company and a project will always follow a certain structure. For example, in the EMEA region, in the HR department, there could be a project about predicting the need for hiring. In this case, including metadata from the existing structure would be important for reporting, monitoring, cleanup, and rollouts. At the same time, the technical department understands that the project will need the following:

• Phases of data collection through a data pipeline that is made up of data ingestion, cleaning and processing

• An ML team to do data modeling for forecasting

• A DevOps pipeline for code orchestration, spread through a dev, test, and prod environment

All of the italicized keywords are business and technical group structures that are important to be associated with the components of an application. This is an example of a typical tagging ontology. Using the example, the following table shows the corresponding key-value pairs for the tags.

Key	Values
department	human resources
region	EMEA
project	hiring forecast
phase	3
process	data ingestion, data cleaning, processing, modeling, or sales forecasting
domain	machine learning or data pipeline
creation	cdk, x framework, ingest pipeline, or manual - empty
status	development, testing, production access, reporting, or onboarding

Governance of tags

Setting up governance mechanisms helps make tagging consistent and programmable across all AWS resources:

• Reactive governance means finding resources that are not properly tagged. You can use tools such as the Resource Groups Tagging API, AWS Config rules, and custom scripts.

• Proactive governance means that users are not allowed to create any untagged resources. You can use tools such as AWS CloudFormation, AWS Service Catalog, tag policies in AWS Organizations, or IAM resource-level permissions.