Warehouse modernization conversations usually start with the database: which cloud platform, which table format, how to convert the SQL. Yet in many organizations the hardest and most expensive part of the estate is the ETL layer that feeds the warehouse: thousands of Informatica PowerCenter mappings, SQL Server Integration Services (SSIS) packages, DataStage jobs or hand-built scripts that encode decades of business rules.
In 2026 two of the most widely deployed ETL tools reached turning points. Informatica PowerCenter 10.5 left standard support, and Informatica itself became part of Salesforce. Microsoft shipped SSIS 2025 while signaling that the future of SSIS workloads lies increasingly in Microsoft Fabric. For data leaders, this is the moment to make deliberate decisions about the ETL estate rather than letting deadlines make them. This guide compares the options, offers decision criteria and walks through a practical approach.
What Changed in 2025–2026
Informatica PowerCenter
According to Tarento’s 2026 analysis of the PowerCenter roadmap, standard support for PowerCenter 10.5.x ended on March 31, 2026, with extended support running through March 31, 2027 and sustaining support beyond that. Informatica is steering customers toward its cloud platform, Intelligent Data Management Cloud (IDMC), which uses consumption-based pricing. The same analysis notes that Salesforce completed its roughly 8 billion dollar acquisition of Informatica in November 2025, with Salesforce emphasizing the role of Informatica’s data management capabilities in its AI agent platform. Check your own contract and Informatica’s official support notices for the dates that apply to your licenses.
SQL Server Integration Services
SSIS remains supported and was updated alongside SQL Server 2025. Microsoft’s Fabric blog describes SSIS 2025 as aligned with the SQL Server 2025 engine, with modern client libraries supporting TLS 1.3 and Microsoft Entra ID authentication, and the ability to target Fabric Data Warehouse. It also notes that the Microsoft Connector for Oracle and the CDC components were moved out of SQL Server 2025 Integration Services, with Fabric Data Factory suggested as the path forward for those scenarios.
In March 2026 Microsoft announced an Invoke SSIS Package activity in preview for Fabric Data Pipelines, allowing existing .dtsx packages stored in OneLake to run from Fabric without a rewrite or a separately managed integration runtime. Microsoft describes this as the first step in bringing SSIS into Fabric, with on-premises connectivity and custom component support planned for later previews.
Support deadlines are a trigger, not a strategy. Use them to sort your ETL estate into what to retire, what to move as-is, and what to rebuild properly, workload by workload.
The Four Paths
Most organizations choose among four broad paths, often combining them across different parts of the estate.
Path 1: Extend and stabilize
Stay on the current tool, buying extended support or upgrading within the same product line, while planning a later move.
- When it fits: a short bridge while a funded migration is underway, or for systems scheduled for retirement soon.
- Risks: rising support costs, a shrinking skills pool, accumulating technical debt and security exposure on aging infrastructure.
Path 2: Move to the vendor’s cloud platform
Migrate PowerCenter workloads to Informatica IDMC, or SSIS packages to Azure Data Factory or Fabric Data Factory, using vendor tooling that preserves much of the existing logic.
- When it fits: organizations committed to the vendor for the long term, with heavy investment in its data quality, governance or master data capabilities, or with a large estate of packages that work well.
- Risks: consumption-based pricing that can grow with data volumes, as Tarento’s analysis warns for IDMC; dependence on preview features in the case of SSIS in Fabric; legacy design patterns carried into a new platform.
Path 3: Re-platform onto the target data platform (lift and shift)
Convert existing ETL logic, largely as-is, into code or pipelines running on the target platform, such as Databricks, Snowflake, BigQuery or Fabric, often with automated conversion tools.
- When it fits: when speed matters, logic is sound and the main goal is exiting the legacy tool and infrastructure.
- Risks: carrying technical debt forward; procedural, row-by-row patterns that perform poorly on cloud engines; and, as Tarento notes, the likelihood of a second clean-up project later.
Path 4: Modernize during migration (rebuild)
Redesign pipelines using native patterns on the target platform: ELT with SQL transformations in tools such as dbt or platform-native pipelines, CDC-based ingestion, declarative orchestration, data contracts and tests.
- When it fits: for core, frequently changing pipelines where maintainability, performance and AI readiness matter, and where the organization can invest more up front.
- Risks: higher initial effort; the need for strong business rule documentation and validation; scope creep if redesign is not disciplined.
Decision Criteria
Rather than choosing one path for the whole estate, evaluate workloads or domains against consistent criteria.
- Business criticality and change rate. Pipelines that feed regulatory reporting or change frequently justify rebuilding; stable, low-value pipelines may be moved as-is or retired.
- Usage. Use run logs and lineage to find jobs that are unused or feed unused reports. Retiring them is the cheapest migration of all.
- Complexity and quality of logic. Well-structured mappings convert cleanly; deeply nested, undocumented logic is a candidate for redesign with business validation.
- Target platform strategy. If your warehouse is moving to Fabric, SSIS in Fabric or Fabric Data Factory may be natural. If you are moving to Databricks or Snowflake, native ELT often fits better.
- Vendor relationship. Consider how much you rely on the vendor’s wider platform for data quality, MDM and governance, and how ownership changes, such as Informatica’s move into Salesforce, affect your strategy.
- Cost model. Compare license, consumption and compute costs over three to five years, including the cost of skills and support.
- Skills. Assess your team’s current skills and the market for them. SQL- and Python-based ELT skills are widely available; legacy tool skills are becoming scarcer.
- Timeline pressure. Support deadlines, data center exits and license renewals constrain how much redesign is feasible.
A Practical Approach: Triage, Then Execute in Waves
- Inventory the estate. Extract metadata for every mapping, workflow, package and job, including sources, targets, schedules, run history and dependencies. Most ETL tools expose this through repositories or export formats.
- Map lineage to consumers. Link pipelines to the tables, reports and downstream systems they feed, and identify owners.
- Triage into dispositions. Classify each pipeline as retire, extend temporarily, move to vendor cloud, re-platform or rebuild, using the criteria above.
- Choose target patterns. Define standard patterns for ingestion (managed connectors, CDC, zero-ETL), transformation (SQL-based ELT, notebooks), orchestration and testing on the target platform.
- Pilot each disposition. Run a small wave for each path to validate tooling, estimate effort and refine patterns.
- Execute in domain-based waves. Migrate by business domain so that owners can validate complete data flows rather than isolated jobs.
- Validate rigorously. Compare outputs between old and new pipelines with automated data diffs and agreed tolerances before cutover.
- Decommission. Turn off legacy jobs and infrastructure promptly after cutover, or savings never materialize.
The Role of AI-Assisted Conversion
AI-assisted tools can now read PowerCenter mapping exports and SSIS package definitions, explain the logic in plain language, and generate equivalent SQL, Python or pipeline definitions. This is useful in several ways:
- Documentation: generating readable descriptions of undocumented legacy logic for business review.
- Conversion acceleration: producing first drafts of transformations for engineers to review and test.
- Pattern detection: identifying repeated patterns that can be replaced with a single reusable component on the target platform.
- Test generation: proposing test cases and validation queries based on the logic.
AI-generated conversions still need human review and data-level validation. Treat them as a productivity tool within a governed process, not as a replacement for engineering judgment.
Building the Business Case
ETL modernization competes for budget with other data and AI initiatives. A credible business case avoids vague promises and compares realistic scenarios over a multi-year horizon.
- Baseline cost of staying: current license and support fees, expected extended support pricing, infrastructure and data center costs, and the staff time spent maintaining legacy jobs and handling failures.
- Migration cost by path: effort estimates from pilot waves for each disposition, tooling and conversion costs, parallel-run costs and validation effort.
- Target run cost: platform compute or consumption pricing at realistic volumes, plus orchestration, monitoring and support. Model growth scenarios, especially for consumption-priced services.
- Risk reduction: exposure from running unsupported software, dependence on scarce skills and audit findings related to legacy controls.
- Enablement value: faster delivery of new data products, fresher data through CDC, and readiness for AI use cases that need governed, well-documented pipelines. Describe these concretely, with named initiatives that depend on them.
Present ranges rather than single numbers, and update the case after pilot waves when estimates become more reliable. Decision makers are more likely to fund a phased program with clear checkpoints than a single large commitment.
Governance and Operating Model After Migration
A modernized ETL layer should not become tomorrow’s legacy. Put basic disciplines in place as part of the program: version control and code review for all pipelines, automated tests and data quality checks in CI/CD, consistent naming and documentation standards, lineage captured automatically from the platform, and clear ownership of every pipeline by a team rather than an individual. Define how new pipelines are requested, built and retired, and review the estate periodically to remove pipelines that are no longer used. These practices keep the benefits of modernization from eroding over time.
Pitfalls to Avoid
- Migrating everything. Without usage analysis, teams spend effort moving jobs that nobody needs.
- One path for all workloads. A blanket decision to lift and shift or rebuild everything rarely fits a diverse estate.
- Ignoring hidden logic. Business rules hide in pre- and post-session SQL, stored procedures, parameter files, scripts and scheduler configurations, not only in mappings.
- Underestimating orchestration. Dependencies, schedules, restart logic and alerting need to be redesigned on the target, not just the transformations.
- Relying on preview features for critical paths. Preview capabilities, such as SSIS in Fabric at the time of writing, are useful for pilots but carry change risk for production-critical workloads.
- Leaving decommissioning for later. Running two ETL estates indefinitely doubles cost and confusion.
A Hypothetical Worked Example
A hypothetical manufacturing group runs about 2,500 PowerCenter workflows feeding an on-premises warehouse, plus several hundred SSIS packages in regional subsidiaries. It is moving its warehouse to a cloud lakehouse and faces the PowerCenter extended support deadline in March 2027.
- An inventory and run-history analysis shows that roughly a third of PowerCenter workflows have not run successfully in a year or feed reports with no recent users. These are retired after owner confirmation.
- Extended support is purchased as a bridge to March 2027, with a firm migration plan.
- Core finance and supply chain pipelines, which change frequently and feed regulatory and management reporting, are rebuilt as SQL-based ELT with CDC ingestion and automated tests.
- Stable, simple workflows are converted with automated tooling onto the lakehouse with minimal redesign.
- Subsidiary SSIS packages are evaluated against the group’s target platform. Since the group is not standardizing on Fabric, packages are re-platformed into the group’s common pipeline framework over time, while a small number continue on SQL Server until their source systems are replaced.
- Each domain wave is validated with automated data comparisons and signed off by business owners before legacy workflows are switched off.
The example is illustrative; the right mix depends on your estate, platform strategy and constraints.
Outlook
ETL is converging into data platforms. Informatica’s future is now tied to Salesforce’s AI strategy, Microsoft is folding SSIS into Fabric, and cloud data platforms increasingly provide native ingestion, transformation and orchestration. Expect AI-assisted conversion to keep improving, managed connectors and CDC to replace more custom extraction logic, and declarative pipeline frameworks to become the default for new work. Standalone, on-premises ETL servers will become rarer each year. Organizations that use the current deadlines to rationalize and modernize, rather than simply relocate, their ETL estates will have simpler, cheaper and more adaptable data platforms by 2028.
Conclusion
PowerCenter’s support timeline and SSIS’s move toward Fabric make ETL modernization unavoidable for many organizations. The best response is neither panic nor a single blanket decision. Inventory the estate, retire what is unused, bridge where necessary, move stable workloads efficiently, and rebuild the pipelines that matter most using modern, testable patterns. Validate every wave and decommission promptly. Done well, the ETL migration becomes the part of warehouse modernization that delivers the most lasting value.
How Hanoi Technology Can Help
Hanoi Technology helps organizations inventory, triage and modernize PowerCenter, SSIS and other legacy ETL estates as part of warehouse modernization programs.