What’s new in SQL Server 2025 CTP 2.0

Three years ago, when the first public preview of SQL Server 2022 (CTP 2.0) was announced, I was a few months in at the SQL Docs team, and had very little to do with that release.

Three years later, the team is slightly larger (we’re called Data Docs now), and I was much more involved with helping scores of people merge the content for SQL Server 2025 (CTP 2.0).

How much effort was the documentation update?

By the numbers for the sql-docs repository (just one of several repositories we manage):

• Over 1,250 Git commits (which we squashed)
• Over 600 files changed (including more than 120 new files)
• Around 50 unique contributors

New administrator features

The features in the first public preview of SQL Server 2025 (CTP 2.0) that I’m most interested in are:

1. tmpfs support for tempdb in SQL Server on Linux. This is the technical way of saying “run tempdb on a RAM drive”. For production workloads, that’s a big deal. And, for non-production workloads (I’m thinking developer containers), there’s nothing stopping you from running all your CI/CD databases in memory.
2. Standard Developer edition. Finally, am I right? MVPs and customers alike have been asking Microsoft for decades to provide a development environment that matches the more affordable SQL Server Standard edition, and now it’s here with SQL Server 2025 (CTP 2.0).
3. Optimized locking. This feature was introduced in Azure SQL a few months ago, and was an obvious upgrade for SQL Server.
4. Accelerated database recovery in tempdb. Another finally! Turn this on with optimized locking, along with tempdb on tmpfs, and you’re unlocking two of the biggest performance bottlenecks in SQL Server today.
5. Remove memory-optimized filegroups. I know this is a little thing, but it’s nice to know you can clean up your files if you’re not using In-Memory OLTP.

New developer features

There are also some developer features that look pretty interesting to me:

1. JSON data type and new functions! This isn’t your grandmother’s NVARCHAR(MAX), but a proper native binary JSON data type. I might have something to say about this soon, as part of my “How data types are stored” series.
2. Improvements to sp_executesql to reduce compilation storms. This happens when lots of queries compile at the same time, which is common on systems that run dynamic Transact-SQL.
3. Fuzzy string matching. SQL Server 2025 (CTP 2.0) introduces four functions to help you calculate the difference between two strings. I could have used this for pattern matching stuff I wrote 10 years ago in CLR code.
4. Regular expressions. Interesting? Definitely. Powerful? Sure. Just don’t expect this to be fast.

I think this is a strong release, and I haven’t even mentioned a lot of the stuff that excites other people, like vectors, Query Store improvements with intelligent query processing, performance improvements for Halloween protection, and support for Base64 encoding and decoding.

Read more about what’s new in this release in the official documentation.