TDM Approaches

Published date: April 15, 2024, Version: 1.0

There are differnt types of TDM Approaches and Best Practces that can be followed, Some of them are:

Data Provisioning Approach

_{Data Masking}

Data Masking creates structurally similar but dummy data that can be used for QE and development activities in non-production environments. Having Data Masked data in a non-production environment will prevent the exposure of sensitive data. IT environments have used production data for testing, development, and enhancement. This poses a risk of customers’ data being exposed. To comply with mandates, it must mask the production data used in the test and development environments. The TDM team should perform Data Masking before the masked data is provisioned/utilized by the QE team.

a) Data elements to be masked

Element-wise sensitivity should cover all the required fields for all the databases across the organization. These elements should be identified and verified by SME / IT security team. The TDM team should provide masking solutions to de-identify the data elements based on the pre-approved data masking rule by the SME.

b) Methodology for masking projects

Masking projects conducted by the central TDM for the data stores should follow the methodology described in this section. The method consists of 4 phases. This same process applies to data subsets and data generation based on their respective areas of requirement.

Scoping: Finalize the list of applications in scope.
Point of contact: Identify the point of contact (QE/Technical/Owner/DBA/Environment Team).
Project objectives: Defines the data requirements and success criteria.
The Environment needs Environment requirements and dependencies

Analyze: Analyze end-to-end data flow for dependencies and referential integrities between tables and their related columns
Data targeted for masking: Identification of the data to be masked, their sizes, complexity, and the actual volume of test data to be masked
Data targeted for masking: Identification of the data to be masked, their sizes, complexity, and the actual volume of test data to be masked
Technical Approach: How the masking will be accomplished against each identified data store, including process flows
Physical Architecture: Representation of the physical data flow regarding environments and masking points
Dependencies: Dependencies for delivery from groups outside of TDM
Operations Plan: Deployment plan into the environments and run schedules

Connection Setup: Set up target data store (database/schema) connections and import metadata
Map and Mask: This phase includes deliverables like mapping documents for masking rules and design executables to conduct the data masking
Document Columns: Create a mapping sheet to know what columns should get masked with the masking rule. These rules will be segregated based on re-usable (global) and project-specific (local)
Design: Design masking rule
Configure: Grouping of tables to mask data. Here, dependency is all the referential integrity between tables DBA shares and the data provider understands. After this, develop a strategy to pull the table to mask in groups

Execution: Execute the implementation according to the configuration plan where table groups are created according to referential integrity
Transition: Transition the “Run” knowledge and turn it over to BAU

Collect database-level referential integrity with the help of DBA
Collect the data masking requirement per table per column
Finalize the business rule per column
Signoff document from SME before proceeding to the configuration of masking rules
Develop and apply the masking rules to the required columns

The diagram below shows the approach to data masking:

Quick Links

System Testing Unit Testing Smoke Test Requirement Analysis

Below mentioned masking types are available:

Key Masking: Masks the key value and produces deterministic results for input key value, masking rule and seed value.
Substitution Masking: Replaces the input column data with a value from a dictionary file or table
Dependent Masking: Replaces the value of the input column based on the value of another column
Random Masking: Replaces input column value with random values for the same source data and masking rule
Expression Masking: Uses Informatica transformation functions to mask and replace data in source columns
Special Mask formats: Replaces realistic fake values for fields like credit cards, addresses, social security numbers etc.
No Masking: The default is no masking.

Data Subset

Data sub-setting creates a smaller, referentially correct copy of a larger database. After subsetting, the cut-down database remains perfectly usable. The data is referentially correct and internally consistent. At this stage, the target environment’s total size is much more manageable.

Subsetting ensures extracting small, more intelligent subsets from production. This allows TDM to quickly provide teams with more manageable sets of consistent, referentially intact data for testing. It also minimizes the risk of exposing sensitive records.

The data can be sub-set based on functional needs by applying data filters. It will be built based on the physical grouping of application tables. The diagram below shows the approach to sub-set the data.

Core requirements for subset criteria should be driven by below factors:

Infrastructure (Space) constraints and costs while managing test environments
The right sample of test data is available for all scenarios to make sure maximum test data coverage

The key steps involved are outlined in the table below:

Operating Systems	Supported Operating Systems
Supported Data Sources	Ex: Oracle 11g, IBM DB2400 iSeries V7R1 etc.
Installation Considerations	TDM Subset can reside on the same system, or does it require any other components.
Subset Procedure
To access the larger data set stored in the data source and apply subset rules to extract small, more intelligent data subsets	Establish database connection(s)
	Create extract definitions
	(Optional) Prepare subset schema
	Create a load group to move data based on driving tables
	Run extracts

Collect database-level referential integrity with the help of DBA
Collect driving tables and the group of tables for which data needs to be filtered
Collect the data subset requirement per table
Finalize the data filter rule per column
Signoff document from SME before proceeding to the configuration of filter rules
Develop and apply the filter rules to the required columns

Eliminate the risk of data breach by creating production-like data without sensitive content
Reduce the infrastructure footprint by increasing test data coverage
Enhance existing subsets of production data with rich & sophisticated sets of synthetic data
Create large volumes of data for non-functional testing like performance and load tests
Simulate wide varieties of data scenarios that include data scenarios not currently available in production
Synthetic data generation services are ideal for transactional data to provision quicker & reliable datasets

Following synthetic data generation techniques may be explored and considered for various test data use cases:

a) Data Generation Techniques

Data Model-based generation: This approach allows users to create complex datasets based on business rules and constraints. Developers build/replicate data models and configure generation rules for each data domain.

Users can assign data elements with the respective domain so that the appropriate generation rule is called out at run time. The most widely used generation rules are:

Random
Lookup
Regular Expressions

Synthetic data is generated at the data source level.

Automated generation via application interfaces: This is a widely used data generation technique. It is based on the data injection via application interfaces (UI) or passing parameters to API to suit the test data requirement.

Synthetic data is keyed into the application interfaces via GUI or API. The process is automated so that manual intervention is reduced.

The approach to delivering synthetic data follows the fulfilment process below:

Activity	Details
Assess Test Data generation requirements	Collect details on master and transaction table relationships Finalize the data business rule per column with the help of SME Finalize the number of records to be generated Verify if the search/reserve service can be leveraged for the requirement Verify if the existing data can be conditioned to fulfil the requirement
Master data vs Transactional data	Identify and categorize generation services based on master data vs transactional
Seed file	Create the seed files for the data dedicated to enterprise level (e.g., Region, Country, type of business product etc.) data set and can be used repeatedly due to its use in multiple tables.
Document signoff	This document has details on each table for which data is to be generated. A tabular display should have details on Table Columns Data type Data length Rules to be applied per column Number of records to be generated for each table

Summarized activities to be followed: Below activities should be followed as a part of the test data generation activity:

Collect database-level referential integrity with the help of DBA.
Collect the data requirement per table per column.
Finalize the business rule per column.
Finalize the number of records to be generated.
Sign off a document from SME before configuring data generation business rules.
Seed file creation to baseline client-specific data used repeatedly across tables.
Develop and apply the business rule per column and configure the number of records generated.

The diagram below shows the approach to data generation:

TDM Best Practices

Minimize reliance on test data. Test data requires careful and ongoing maintenance. You must update or re-create related test data as your APIs and interfaces evolve. This process represents a cost that can negatively impact team velocity. Hence, minimizing the amount of test data needed to run automated tests is good practice.
Isolate test data. Run your tests in well-defined environments with controlled inputs and expected outputs that can be compared to actual results. Ensure that data consumed by a particular test is explicitly associated with that test and isn’t modified by other tests or processes. Wherever possible, your tests should create the necessary state as part of the setup using the application’s APIs. Isolating your test data is also a prerequisite for running tests in parallel.
Minimize reliance on test data stored in databases. Maintaining test data stored in databases can be particularly challenging for the following reasons:
- Poor test isolation. Databases store data durably; any changes to the data will persist across tests unless explicitly reset. Less reliable test inputs make test isolation more complex and can prevent parallelization.
- Performance impact. Speed of execution is a key requirement for automated tests. Interacting with a database is typically slower and more cumbersome than interacting with locally stored data. Favour in-memory databases where appropriate.
Make test data readily available. Running tests against a copy of a complete production database introduces risk. It can be challenging and slow to get the data refreshed. As a result, the data can become out of date. Production data can also contain sensitive information. Instead, identify relevant sections of data that the tests require. Export these sections regularly and make them readily available for tests.

Roles & Responsibilities

TDM Team

Role	Responsibility
Test Data Architect	Define Organizational Strategy / Policy covering Test Data Archival, Data Refresh, Data Sub-Setting, Data Generation, Data Masking etc. Work with key stakeholders for TDM process enhancement/refinement Review Data Strategy for Large, Complex Programs Define Enterprise Test Data Tool Strategy Create a Test Data Delivery plan as per Test Data Requirements, Release Plan & Environment Availability TDM Risk Management Timely escalation to QE leadership Test Data Governance & Monitoring
Test Data Analyst	Analyze the Data Requests provided by the Requestors Create Test Data as per specifications – including batch setup as needed. Liaise with Development and Environment team as needed on batch jobs. Create Test Data mapping against TDR. Identify opportunities for automation. Participate in Defect Triaging for Test Data related issues Provide Estimates on Data Requests deliverables Support Creation of Test Data Utilities Support customization and use of CTC’s TDM utilities/assets as applicable.

Role

Responsibility

Test Data Architect

Define Organizational Strategy / Policy covering Test Data Archival, Data Refresh, Data Sub-Setting, Data Generation, Data Masking etc.
Work with key stakeholders for TDM process enhancement/refinement
Review Data Strategy for Large, Complex Programs
Define Enterprise Test Data Tool Strategy
Create a Test Data Delivery plan as per Test Data Requirements, Release Plan & Environment Availability
TDM Risk Management
Timely escalation to QE leadership
Test Data Governance & Monitoring

Test Data Analyst

Analyze the Data Requests provided by the Requestors
Create Test Data as per specifications – including batch setup as needed.
Liaise with Development and Environment team as needed on batch jobs.
Create Test Data mapping against TDR.
Identify opportunities for automation.
Participate in Defect Triaging for Test Data related issues
Provide Estimates on Data Requests deliverables
Support Creation of Test Data Utilities
Support customization and use of CTC’s TDM utilities/assets as applicable.

Key Stakeholders:

Team	Responsibilities
Data Migration Team	Perform Data refresh activities as per schedule or on request.
Development	Request data for development activities.
QE Specialist	Request testing environments to be set up for QE activities. Request data from the TDM team as per functional requirements. Request for data refresh.
Information Security Team	Run Data Masking process as per TDM team request. Create Data Masking Strategy and Masking Solutions for new data/environment as requested by the TDM team.
Release Management	Communicate Release Management Strategy, including Release Cycles, Environment and Path Usage, in advance to the TDM team and other stakeholders.

Entry & Exit Criteria:

Phase	Entry	Exit
Assessment / Intake	Project Charter Stakeholders’ identification from the QE team, Development team, TDM team and others Kick-off meeting	Criticality of project New/existing (transformation / BAU) The scope and Release schedule finalized
Test Planning	High-level Requirements, TDM fulfilment window time / Run plan Reference Strategy, Test Data plan Slice details	Test Data Estimates, Data Request Test Data Plan, Slice finalized Users, Access privileges to the Source and Target environments
Test Design	Test Data Requirement Data Request ALM project setup Source / Target environments / Files	Data Scripts (as-is / created/modified) Run data scripts for sanity, regression testing Data Masking done Test Data Requirement Traceability document up to date No outstanding data issues Test Data Provisioning Summary Report (TDSR) published The lessons learnt paper goes into the Data Assets repository QE team sign off on Test data fulfilment
Test Execution	Provisioned Test Data for the given scope end to end Additional test data requests Data Conditioning (if any)	BAU support Test Data set (snapshots if required) Data purge/intimations to the Environment

Suspension & Resumption Criteria

Suspension Criteria

All test data provisioning efforts should adhere to the scheduled timelines unless one of the following critical dependencies is altered.

There are significant bottlenecks/showstoppers for which the test data provisioning activity cannot be continued.
In case of Known / Unknown test environment downtime.
External dependency on the project is unable to provide required triggers/data inputs may cause halting the test data activities. For example, a) SME availability to provide the data model or clarify a critical query on a data model, b) Source environment shut down.
Hardware/software not available at the time indicated in the project schedule.
Any unplanned holiday shuts down both developments (IT) and testing.

Resumption Criteria

The data provisioning activity will be resumed once the bottlenecks/showstoppers issues are settled.
Once the test environment is up, the test data activity will be resumed.
Test data provisioning will be resumed once the TDM team receives the required triggers/data inputs.
Test data provisioning will be resumed once the necessary hardware/software is available.