SQL Server Integration Services still ships with SQL Server 2025 [5], so this is a lifecycle question. SQL Server 2016 left extended support on July 14, 2026 [1]; SQL Server 2017 leaves it on October 12, 2027 [2]; SSIS 2025 removed the Attunity change-capture components, the Microsoft Connector for Oracle and the Hadoop tasks [6]; and running packages in Azure means paying for an SSIS Integration Runtime by the node-hour [7]. Lima converts packages in batches: a package becomes a pipeline, a data flow becomes operators, a script task becomes Python, the removed change-capture components become Lima's change capture, and SQL Agent jobs become schedules with run windows.
Nothing forces a package to stop on a support date. What changes is that the SQL Server it runs on stops receiving fixes, the change-capture components you may depend on are gone from the current release, and the cloud routes for SSIS come with meters and limits. Estates on 2019 and 2022 have years; estates on 2016 and 2017 have a decision.
The conversion is a mapping, and the mapping is short. What has no equivalent is listed too.
| Today | In Lima | What to know |
|---|---|---|
| Package (.dtsx) | Pipeline on the canvas | One package, one pipeline; the inventory groups packages by pattern. |
| Control flow (tasks, precedence constraints) | DAG of steps | Sequence, success and failure paths carry over as dependencies. |
| Data flow (source, transformations, destination) | Operators | Derived Column, Lookup, Conditional Split, Aggregate, Sort, Merge Join, Union All have direct operators. |
| Lookup with cache | Cached lookup query | Same semantics; the database caches the answers. |
| Slowly Changing Dimension | Upsert and history load | Type 1 becomes an upsert; type 2 becomes a history table with validity dates. |
| Script Task / Script Component (C#, VB) | Python operator | Logic is rewritten in Python; the inventory sizes each script. |
| Attunity CDC components (removed in SSIS 2025) | Lima change capture | SQL Server CDC tables or change tracking; Oracle LogMiner where the DBA allows; Lima's own checkpoints. |
| Microsoft Connector for Oracle (removed in SSIS 2025) | Oracle connector | Partitioned JDBC reads within an agreed cap. |
| Error output (redirect row) | Exception records with owners | Bad rows kept and routed with a deadline; replayed once fixed. |
| Project parameters and environments | Connection registry and pipeline parameters | Environment values live with the connection or the schedule. |
| SQL Server Agent jobs | Orchestration on Airflow | Schedules, dependencies, run windows with deadlines; an enterprise scheduler can trigger Lima instead. |
| SSISDB catalog reports | Run reports and lineage | A reconciliation report per run; lineage per number. |
The assessment reads exported projects and the catalog's execution history, and needs no access to production data.
| What it reads | What it reports |
|---|---|
| Packages and projects | Counts by pattern: copies, lookups and joins, SCD loads, scripts, change capture. |
| Components by family | Sources and destinations with the Lima connector for each; removed components (Attunity CDC, Oracle connector) flagged. |
| Scripts | Script tasks and components by size and language, with the Python conversion effort. |
| Execution history | Run times and failure rates from SSISDB, so run windows can be set. |
| Effort bands | Per batch of ten to twenty packages. |
One week. You send the exported projects and a catalog export; we return the count by pattern, the removed-component flags and the effort bands. No charge.
Get the assessment →Fixed priceThe first ten to twenty packages converted, run side by side, reconciled and cut over on a fixed price. Packages using removed components first.
Offers and principles →SQL Server stays: as a source, as a target, or both. Lima reads it in partitions and captures changes through its CDC tables or change tracking.
SQL Server Agent or your enterprise scheduler can keep triggering jobs while the packages move; Lima's external-scheduler trigger reports status back.
You have until January 2030 on extended support [3]. The reasons to move earlier are the removed change-capture and Oracle components in the current release, the cost of running packages in Azure, and the checks, owners and reports a Lima pipeline adds. The inventory gives you the effort figure.
Microsoft's 'Invoke SSIS Package' activity in Fabric Data Factory is in preview with a Fabric-managed runtime; on the check date it did not support on-premise sources or custom components [8].
They are rewritten as Python operators. The inventory lists each one with its size; most are short data-cleaning routines.
One week, no charge. You get the count by pattern, the exposure flags and the effort bands, whatever you decide next.
