Organizations with cutting-edge DevOps tools and practices still find it difficult to provision right data at the right time for continuous testing framework. Test Data Management (TDM) is a pre-requisite for the success of DevOps and Continuous integration/Continuous delivery (CI/CD)  methodology.

TDM is no longer a traditional support and back office function, but a core and critical function that acts as a business enabler for security, agility and cost efficiency. There is a need for new TDM methodologies and strategies supported by efficient tooling strategy that can solve data provisioning challenges and help adopt DevOps and CI/CD methodologies.

How organizations manage test data today?

Organizations manage test data in multiple ways, right from maintaining them in simple excel sheet or CSV file to temporary database on shared environment:

• Shared test data environments - Testers share test data sets among their peers and teams in shared environments. This provides easy maintenance and access of test data to project teams, but creates challenges wherein one stakeholder has a conflict with another for the same record or same resource at the same time.
• Synthetic data creation - Create/generate data based on data models that resemble the production data. As a substitute to production data, it can do away with the process of subset and masking. This is a risk-free approach as it ensures security of personal information or real data or potentially sensitive data.
• Sub setting logics – This is used to limit the production copy of entire data set to limited set, enabling faster, quicker copy of data. However, the challenge lies in identifying the right subset that can provide 100% test coverage.
• Isolated masking solutions – This is used to ensure that the personal identifiable data are desensitized complying to PII, PHI, HIPAA and country laws before landing into test environments.

Test data management has evolved over the past few years.  Figure 1 illustrates the different phases of progression in test data management. 

Early TDM approach was project-centric applications. This repurposed DWH/Dimension tables to serve as a test data repository with lack of focus on quality and governance of test data. The next phase, TDM 1.0 was designed to be more aligned to domain-centric approach i.e. Department/BU focused projects/local applications with awareness about data quality and reactive data governance. It was mostly focused on a domain and was more of an IT centric Initiative.

TDM 2.0 increased focus on data governance with focus on stewardship and business workflows besides core TDM functionalities. Interest on multi-domain TDM started emerging and there was increased business and IT alignment on TDM. With TDM 3.0, the overall TDM ownership and interest shifted towards the business stakeholder. TDM became proactive data governance and the overall architecture maturity shifted from consolidation to centralization.

TDM 4.0 and beyond

The current trend, TDM 4.0, is the expansion and optimization phase, which includes enabling DevOps, IoT, etc. In this phase, Machine Learning and Big Data will drive TDM strategy and the rise of Chief Data Officer in enterprises will mean that data management will have greater visibility and transparency in the enterprise data initiatives supported by new technologies like Database Virtualization, Service Virtualization, and Containers.

TDM strategy for DevOps and other emerging technologies should be built on the approach that we call the 4 Dollar, 3 Penny (4D,3P) principle (See Figure 2).

Along with the 4 Dollar, 3 Penny principle, the below given tenets should also be implemented for a good TDM strategy:

Secure at source
Data copied from production should be masked before landing in non-prod environments. On the fly masking should be adopted complying with regulatory and local laws. There are multiple TDM tools that adopt this process.

DevOps, Agile adaptability
Test data creation and provision should support DevOps and Agile methodologies leveraging test-driven data creation frameworks. This requires faster turnaround of data refresh, masking and creating ad hoc data requests to support ongoing testing activities.

Self service
Data request should be catered to all stakeholders as on-demand service delivery. The stakeholders can be developers, users, testers, data base administrators, data SME. This self service should help to reduce the turnaround time for refreshing the entire data/selected data set, provision only masked data, do periodic refreshes to golden copy, archive and backup the data. All these services should be provided as a service and catalogue based delivery model.

Synthetic data adoption
Programs should adopt ways and means to create dynamic data or synthetic data to cover out layered scenarios or negative scenarios that cannot be tested using existing production data. These synthetic data can be created using licensed tool/open source tools/simple scripts using .Net/Java platform.

Support to all testing phases
Test data should cater to all types of testing including functional, regression, automation, performance and user acceptance testing. Framework should support the development team to populate required data to unit test environments as well.

The future of test data management
Test data is traditionally dependent on production data with exceptions to use sensitized data in test environment. With improvements in technology, we see many open source companies and third party tool vendors adopting synthetic data, as it is more convenient and useful to create data for Agile/DevOps programs. Service virtualization supports creation of test stubs and complements TDM tools. Dependency on production data should be reduced as it involves huge turnaround time including masking sensitive data. The overall focus should be on leveraging the power of synthetic data generation, supplemented with service virtualization methodologies. We can still leverage production data to replicate any critical scenarios that need to be recreated in lower environments.
 

Sathiya Narayanan G

Senior Consulting Manager - Data Assurance, Wipro

Sathiya has over 18 years of experience in the IT industry with technical expertise on Data Warehouse, Test Data Management, and Big data systems. He is PMP and OCP certified, and is currently working on providing solutions on Data, Environment and Cloud assurance platforms to global enterprises.