New ZSTD Backup Compression Algorithm in SQL Server 2025
With the arrival of SQL Server 2025, Microsoft has introduced ZSTD as a new backup compression algorithm. This addition aims to significantly optimize backup operations, improving both execution speed and efficiency in the use of system resources.
The introduction of ZSTD in SQL Server 2025 represents one of the most important changes in backup management in recent years, especially in environments where data volume, recovery times, and storage consumption are critical factors.
Evolution of backup compression in SQL Server
Backup compression before SQL Server 2008
Before the release of SQL Server 2008 Enterprise Edition, backup compression was not natively available in SQL Server. Database administrators were forced to rely on third‑party solutions that provided specific tools for managing and compressing backup files.
MS_XPRESS and native compression
With SQL Server 2008, Microsoft introduced native backup compression for the first time using the MS_XPRESS algorithm. Later, in SQL Server 2012, this functionality was extended to all editions of the database engine, making compression widely available.
QAT in SQL Server 2022
After more than a decade without major changes in this area, SQL Server 2022 introduced the QAT (QuickAssist Technology) algorithm. However, its performance was strongly tied to the use of specific Intel hardware, which limited its real impact in many production environments.
What is ZSTD and why is it key in SQL Server 2025?
What is ZSTD (Zstandard)?
ZSTD, also known as Zstandard, is a fast, open‑source, lossless compression algorithm designed to provide an optimal balance between compression speed and compression ratio. Its adoption in SQL Server 2025 marks a turning point in the product’s backup strategy.
Key advantages of ZSTD
- Higher compression ratio than traditional algorithms such as MS_XPRESS.
- High decompression speed, ideal for scenarios with frequent restores.
- Scalability through different compression levels.
- Performance independent of the hardware manufacturer.
How to use ZSTD in SQL Server 2025
Basic backup syntax with ZSTD
To use ZSTD for backup compression, simply add the parameter
WITH COMPRESSION (ALGORITHM = ZSTD)
to the BACKUP DATABASE command.
Available compression levels
SQL Server allows you to define the compression level using the LEVEL parameter, which can take the values LOW, MEDIUM, or HIGH. The higher the compression level, the higher the CPU usage and the longer the time required to complete the backup. If no level is specified, LOW will be used by default.
Example of a backup with ZSTD and MEDIUM level
BACKUP DATABASE test TO DISK = 'C:\BACKUP\test-ZSTD-medium.bak' WITH COMPRESSION (ALGORITHM = ZSTD, LEVEL = MEDIUM);
Real‑world ZSTD tests in SQL Server 2025
Test environment
The tests were carried out using a 444 GB database, comparing different compression algorithms and compression levels to evaluate backup times, final file size, and restore times.
| Algorithm | Backup time | Size (GB) | Compression % | Restore time |
|---|---|---|---|---|
| No compression | 00:38:45 | 256 | — | 00:08:22 |
| MS_XPRESS – LOW | 00:27:28 | 128 | 50.39 | 00:09:21 |
| MS_XPRESS – MEDIUM | 00:25:13 | 121 | 53.10 | 00:07:00 |
| MS_XPRESS – HIGH | 00:30:18 | 121 | 53.10 | 00:07:01 |
| ZSTD – LOW | 00:16:15 | 129 | 50.00 | 00:05:22 |
| ZSTD – MEDIUM | 00:28:28 | 106 | 58.91 | 00:06:26 |
| ZSTD – HIGH | 01:06:00 | 100 | 61.24 | 00:06:30 |
Results analysis
The results show that ZSTD delivers clearly superior performance. The ZSTD LOW level achieves the fastest backup, reducing the execution time to less than half compared to an uncompressed backup. On the other hand, ZSTD HIGH provides the highest compression ratio, reducing the backup size by more than 60%, although with a significant increase in execution time.
ZSTD MEDIUM positions itself as the best balance between speed and compression ratio, achieving 9% more compression than LOW while requiring less than half the time needed by HIGH.
Comparison with MS_XPRESS
When directly comparing MS_XPRESS with ZSTD across different levels, ZSTD delivers better results in all scenarios. Even at the HIGH level, although backup time increases, the improvement in compression ratio can be decisive in environments where storage space is a critical factor.
CPU usage with ZSTD
In additional tests performed on a database of approximately 8 TB, CPU usage for ZSTD LOW is practically identical to MS_XPRESS. As the compression level increases, CPU usage rises progressively during both backup and restore operations.
| Algorithm | Backup CPU | Restore CPU |
|---|---|---|
| MS_XPRESS | 13% | 6% |
| ZSTD – LOW | 13% | 6% |
| ZSTD – MEDIUM | 14% | 7% |
| ZSTD – HIGH | 16% | 8% |
Conclusion
ZSTD in SQL Server 2025 represents a significant evolution in backup compression. Its different compression levels make it possible to adapt backup strategies to the needs of each environment, whether prioritizing execution time, storage savings, or reduced recovery times.
If you want to take full advantage of ZSTD and the new features in SQL Server 2025, feel free to contact us to help you with your migration and optimization process.
Additional information
BACKUP (Transact-SQL) – SQL Server | Microsoft Learn
ZSTD compression in SQL Server 2025 | Microsoft Community Hub
SQL Server 2025 : Zstandard Backup Compression – Simple Talk
SQL Server 2025 Zstandard (ZSTD) Backup Compression: Expert Guide
SQL Server 2025: Using ZSTD Compression for SQL Server Backups
SQL Server 2025 Backup Compression and Restore Review
Database administrator with several years of experience in database monitoring, backed by a higher-level qualification in network systems administration. My focus is always on continuous improvement and exploring new technological solutions. I strive to stay up-to-date with the latest innovations and never settle for the status quo; I am always eager to explore and learn new technologies to continue growing both professionally and personally.
