SSRS is gone. SQL Server 2025 shipped without it — the first version to drop SSRS entirely, with Power BI Report Server (PBIRS) as the named replacement. But what about SSIS?
There is no EOL yet, but the signals are clear. SSIS is being phased out in favor of Microsoft Fabric, which is positioned as the next-generation unified analytics platform. Look at that. They don't even call it 'ETL' anymore. Again, it's not gone yet, but Microsoft is encouraging the modernization of ETL workloads to Fabric to address the limitations of on-premise SSIS infrastructure, such as limited scalability and high maintenance overhead.
We've been here before. DTS became SSIS with SQL Server 2005 — Yukon, for those of us who were there. I was. [Thank you, Donald Farmer!] The transition wasn't a cliff either. It was deprecation notices, a long runway, and then one day DTS was just gone. The difference this time is that the runway leads to the cloud.
SSIS 2025 Shipped. Read the Fine Print.
SSIS 2025 is generally available, bundled with SQL Server 2025. Microsoft announced it at Ignite in November 2025 within the Microsoft Fabric Blog, not the SQL Server blog. That's not an accident.
Here's what shipped with SSIS 2025 — the full picture, not just the press release version.
The one new feature:
The ADO.NET connection manager now supports the Microsoft SqlClient Data Provider, with Microsoft Entra ID authentication and TLS 1.3 support. That's it. One new feature in the entire release.
What's deprecated:
Legacy Integration Services Service — the service that lets SSMS monitor packages stored in the SSIS Package Store or msdb. Not a big deal if you've migrated to the SSIS Catalog (SSISDB), but if you're still on the package deployment model, this will matter.
32-bit execution mode. SSMS 21 and SSIS Projects 2022 are 64-bit only going forward. If you have custom or third-party components that are 32-bit only, they break.
SqlClient Data Provider (SDS) connection type. The recommendation is to migrate to the ADO.NET connection type.
What's been removed entirely:
CDC components by Attunity and CDC service for Oracle by Attunity — gone. If you're using these for change data capture from Oracle, you need a replacement. Now.
Microsoft Connector for Oracle — also gone. Support ended July 2025. Not available in SQL Server 2025 onward.
Hadoop components — Hadoop Hive Task, Hadoop Pig Task, Hadoop File System Task — removed.
The breaking change nobody talks about:
The Microsoft.SqlServer.Management.IntegrationServices assembly now depends on Microsoft.Data.SqlClient instead of System.Data.SqlClient. If you have PowerShell scripts, C# automation, or deployment tooling that instantiates the IntegrationServices object, those scripts will break without updates. This one is quiet, easy to miss, and will bite you during your first post-upgrade deployment.
Where Microsoft Is Pointing
The official migration path is Microsoft Fabric Data Factory. SSIS lift-and-shift in Fabric is currently in private preview. That is the ability to run existing SSIS packages inside a Fabric workspace without rewriting them. Microsoft's framing is 'bridge, not cliff,' and that's fair. But a bridge going one direction is still a one-way bridge.
For what it's worth, SQL Server 2022 extended support runs to January 11, 2033, and SSIS goes with it. But that's seven years, not forever, and migrations of complex SSIS estates take longer than people expect.
What You Should Actually Do Right Now
Nobody is forcing you off SSIS today, but here's what's worth doing now, before anything becomes urgent:
Know what you have.
Inventory your SSIS packages. How many are there? Where are they deployed — SSIS Catalog, msdb, file system? Which ones use Oracle connectors, CDC components, or 32-bit custom components? If you can't answer those questions, you can't plan.
Check your automation scripts.
If anything in your environment uses Microsoft.SqlServer.Management.IntegrationServices — PowerShell deployment scripts, CI/CD pipelines, custom monitoring — verify it's referencing Microsoft.Data.SqlClient before you upgrade to SQL Server 2025. This is the most likely silent failure on upgrade day.
Identify your Oracle-dependent packages.
The Microsoft Connector for Oracle is gone in SSIS 2025. If you're upgrading SQL Server and you have packages using that connector, they will not work. Third-party alternatives exist — ADO.NET components, KingswaySoft, and others — but you need to know which packages are affected before upgrade day, not after.
Don't migrate to Fabric yet just because Microsoft says to.
Fabric Data Factory is a genuinely different product. It's cloud-native, SaaS, and not a drop-in replacement for on-premises SSIS. The lift-and-shift option in private preview is promising, but it's still preview. For production ETL workloads, 'let's move to Fabric' is a project, not a setting. Treat it that way.
So, Is SSIS Next?
Brent Ozar asked this question in June 2025. The community answered the SSRS half loudly. The SSIS half got quiet. Here's my answer.
No EOL date has been announced. SSIS 2025 shipped, which means Microsoft is still investing in it — just barely, and clearly with both eyes on Fabric. The trajectory is obvious even if the timeline isn't. SSRS got a clean announcement and a named replacement. SSIS is getting a slower walk toward the exit, with a Fabric-shaped door at the end of the hallway.
I think Brent put it well in his comments: without a fully on-premises replacement for SSIS, Microsoft can't make a clean announcement the way they did with SSRS. The closest thing on-premise shops have is a self-hosted integration runtime — but the control still depends on the cloud.
For now: don't panic, do inventory, review your automation scripts, and watch what happens with SSIS lift-and-shift in Fabric once it goes GA. That's when the real Fabric migration conversation starts. Let me know how I can help.
More to Read:
What's New in Integration Services in SQL Server 2025 — Microsoft Learn
The Evolution of SSIS: SSIS 2025 Generally Available — Microsoft Fabric Blog
SSIS Microsoft Connector for Oracle Deprecation — Microsoft SQL Server Blog
SQL Server 2022 Lifecycle — Microsoft Learn
Rebecca, I don't think we have met but let me introduce myself online. My name is Ivan Peev and I'm one of the founders of COZYROC. We have been producing components for SSIS for almost 20 years now. I can confidently claim we are one of the main SSIS authoritities because we have been longer in this market compared even to the people responsible for SSIS at Microsoft at the moment who have less time spent on SSIS. Brent Ozar is not SSIS authority either, although I respect his DB expertise. I drink, sleep, dream SSIS all day, everyday. That is my livelihood and passion. And I'm deeply knowledgeable about SSIS.
ReplyDeleteSo to the point, SSIS isn't dead and there is no writing on the wall that you speak of. If someone claims, otherwise I suggest he/she reveals and I will challenge all these wrong assumptions. Here are a couple of points for you to chew on:
* It doesn't matter whether Microsoft offers new SSIS features or not. The SSIS third-party ecosystem is so strong that it trounces anything else produced by Microsoft in the integration space. I'm not kidding. SSIS is a monster integration platform.
* CDC Components - COZYROC is currently beta testing replacements for the ;egacy components.
* Connector for Oracle - we have offered one for more than 15 years now.
* Hadoop connectors - nobody cares because nobody uses Hadoop. Everyone is now on Spark.
---
Fabric Data Factory can never be a viable replacement for SSIS because it is : slower, not extensible, missing important enterprise features, only available in the cloud and frankly not properly designed. SSIS will continue to be with us for a very long time with us because its design is timeless and you can constantly build on it further and further. I guarantee you have not seen the last of it yet.
Let's put the grandiose announcements to rest by having a live debate and conversation on the topic of SSIS.
Ivan, welcome — and what an introduction! Twenty years building for SSIS is not a small thing, and COZYROC's component library is something I've pointed customers to more than once. Respect.
ReplyDeleteThe title of the post is SSIS is not dead. Yet. That 'yet' is intentional. I'm not standing at the grave — I'm standing at the window watching the clouds roll in. Honestly, I'd love to be wrong. I love SSIS and I have customers running in production right now and I'd like to see it working for another decade. But the Microsoft signals are hard to ignore. SSIS 2025 shipped with exactly one new feature — ADO.NET support for the new SQL client. That's it. Meanwhile, the list of breaking changes and deprecations is long, the Oracle connector is gone, CDC is gone, and the announcement landed on the Microsoft Fabric Blog, not the SQL Server blog. That was not an accident. Directions on Microsoft put it plainly: SSIS has gone beyond stalling on-premises. Microsoft is actively steering affected customers toward Fabric Data Factory as the answer to those removed components. That's the writing on the wall I'm referring to.
Your points on the third-party ecosystem are well taken, and honestly they underscore my point more than they refute it — the platform is surviving on vendor investment, not Microsoft's. COZYROC stepping in on CDC is a perfect example of exactly that dynamic.
Where I'd push back a little: "Fabric Data Factory can never be a viable replacement" is a confident claim for a product that's still maturing and actively adding SSIS lift-and-shift capabilities. "Not yet viable for production" — Agreed. Never is a long time in this industry, as everyone who said that about the cloud in 2008 would probably tell you.
As for the live debate — I'm flattered, but I write a blog, not a courtroom. I'll keep watching and writing, and I suspect we'll largely be agreeing more than arguing.
This is a great discussion.
ReplyDeleteI think both perspectives (yours, Rebecca, and Ivan's) can be true at the same time.
Microsoft’s direction is clearly toward Fabric.
And the SSIS ecosystem is still very strong.
But in my experience, most SSIS pain doesn’t come from the platform itself.
It comes from lifecycle gaps:
• limited change visibility
• low deployment confidence
• inconsistent execution patterns
• weak governance
That same pattern shows up in ADF and Fabric as well.
Whether teams stay on SSIS or move forward,
lifecycle discipline is the multiplier.
Without it, every platform feels fragile.
With it, even legacy platforms become reliable.
Andy, thank you for that — and yes, I think you're right that both can be true at the same time. That was actually the intent of the post: not 'SSIS is dying tomorrow', but the signals are there and worth becoming aware of. Ivan and I may frame the urgency differently, but I don't think we disagree on the platform's capability.
ReplyDeleteYour point about lifecycle discipline is one I find myself returning to again and again. The shops I've seen struggle most aren't doing so because of SSIS — they're struggling because they're not sure what they have, documentation is lacking, and deployments are combined w/prayers and finger crossing. No judgement intended. And you're absolutely right! That problem doesn't go away with Fabric. It follows us there. Same weakness, just a different flavor.
The inventory question I raised in the post — know what you have, know where it's deployed, know what breaks — I believe that could be the foundation of the lifecycle discipline you're describing. Whether a shop stays on SSIS through 2033 or starts planning a Fabric migration next year, step one is the same.
And same question exists for SSAS
ReplyDeleteI will even say it exists 2 times : SSAS Multidim (yes I am a MDX guy...) and SSAS tabular
As SSAS tabular never completely outperform/cover all features of SSAS multidim they had to maintain both and multiple SAAS rely on it for analytics without cloud fees (on prem or private cloud or VM)
We heard SSAS multidim may be deprecated within 5 years so we start to look for a replacement :
Microsoft has zero alternative not cloud based in fabric, we will look on tools like clickhouse or duckdb
SQL server will continue to evolve but you can bet a lot of stuff will be "fabric" only without technical justification
The parallel is a fair one — and honestly the SSAS Multidim situation might be a starker version of the same story. At least SSIS has a stated lift-and-shift path into Fabric, even if it's still in preview. From what I understand, SSAS Multidim has no equivalent runway being built. Tabular absorbed some of the use cases but not all of them, and you're right that a lot of shops kept both running for exactly that reason. The 'Fabric only without technical justification' framing is something I hear around SSIS too — it's hard to have an honest migration conversation when the answer to every on-prem constraint is 'go to the cloud.' ClickHouse and DuckDB are interesting options worth watching. I haven't covered SSAS much here yet - but not for the lack of things to say. Thank you for the nudge.
Delete" Microsoft is encouraging the modernization of ETL workloads to Fabric to address the limitations of on-premise SSIS infrastructure, such as limited scalability and high maintenance overhead. "
ReplyDeleteSure, but in all honesty: offering a cloud service is simply the more interesting policy to follow. And that is : For Microsoft.
A large range of companies do not need Fabric and smaller companies are simply let to believe they need cloud solutions like Azure databases that are a real big step back compared with traditional SQL server.
Stepping back from lifting everything into the cloud may be a prudent strategy . I'm glad my company does..
ha! 'interesting policy' is a very diplomatic way to put it. You're not wrong — the push toward cloud-first messaging benefits Microsoft's recurring revenue model more than it solves actual problems for the majority of their customers. SSIS running on-prem for a mid-size company with stable, predictable ETL workloads isn't a 'limitation' — it's a norm.
DeleteI believe the scalability and maintenance overhead arguments are real in some contexts, but they're applied as a blanket justification regardless of fit. A company moving a handful of SSIS packages to Fabric isn't modernizing — they're just paying more per month to do the same thing with many extra steps.
Glad to hear your company is keeping its feet on the ground. There's something to be said for simple infrastructure that works. KISS, I believe, is the term, and I support it fully.