Key takeaways
- Recent architectural breakthroughs in the core database engine have unlocked a 2.5x increase in throughput, allowing the software to scale linearly with modern multi-core hardware.
- By removing legacy bottlenecks like centralized serialization, the latest updates ensure high-performance speeds are available out-of-the-box without requiring manual tuning.
- Independent industry benchmarks confirm that the newest versions sustain millions of transactions per minute, proving the software’s readiness for intensive enterprise workloads.
- Evaluating performance based on older versions often misses these critical engine optimizations, making it essential to baseline against current releases for an accurate technical picture.
It is hard to believe it has already been a year since I joined MariaDB as Vice President of Engineering. Looking back at my journey – from resolving customer bugs to leading elite engineering teams – coming to MariaDB felt like a natural next step. My background has always been rooted in building high-performance systems and fostering cultures of engineering excellence, and the mission here was clear: take the world’s most versatile open-source database and make it the fastest, most scalable engine for modern enterprise workloads.
When I stepped into this role, I knew the talent within this organization was world-class. Over the last twelve months, we haven’t just maintained the status quo; we have fundamentally re-engineered the core database engine to meet the demands of today’s high-core, NVMe-driven hardware.
A Year of Breakthroughs: The 2.5x Performance Story
I am incredibly proud to share that our engineering team’s relentless focus on performance has paid off. In October 2025, we published a milestone report: Performance Engineered: MariaDB Enterprise Server 11.8 Accelerates OLTP Workloads by 2.5x.
This wasn’t an incremental “check-the-box” update. For years, database scalability has been hampered by centralized serialization points – bottlenecks that become glaringly obvious on modern hardware. By redesigning LSN (Log Sequence Number) generation and removing critical mutex bottlenecks, the team enabled MariaDB to scale linearly with modern multi-core CPUs. In our testing on Dell PowerEdge R7715 servers with AMD EPYC™ processors, we achieved a 2.5x improvement in OLTP throughput compared to our 10.6 long-term release.
This achievement is a testament to the “performance-first” culture we’ve built. I want to specifically call out the fantastic work from Marko Mäkelä, Rahul Raj and Axel Schwenke. Their deep expertise in InnoDB internals and their ability to analyze and solve complex concurrency issues like the LSN allocation bottleneck (MDEV-21923) and MDL scalability (MDEV-19749) have been nothing short of brilliant.
Precision Matters for Performance Testing
While we celebrate these wins, it is also important to address the nuances of modern database benchmarking. We have noticed some recent public performance benchmarks that unfortunately evaluated older versions of MariaDB that don’t include recent performance enhancements.
This highlights a critical challenge in the ecosystem: when a database is tested without the latest architectural breakthroughs, the results reflect “saturation points” that we have already engineered away. To get an accurate view of modern MariaDB, it is essential to baseline against our latest engine optimizations – otherwise, the data represents a legacy snapshot rather than the current state of the art.
Why Independent, Modern Testing Matters
To get a true sense of where the industry stands, it is vital to look at independent, rigorous performance testing that utilizes modern methodology. This is why the work of Steve Shaw at HammerDB is so important.
Steve’s conclusions demonstrate that when MariaDB is tested against the industry-standard HammerDB TPROC-C benchmark on production-grade infrastructure, it sustains millions of transactions per minute with no manual tuning required. His work confirms that MariaDB Enterprise Server 11.8 (also available in MariaDB Community Server 12.3.0) converts low-level engineering improvements into massive, out-of-the-box scalability for the enterprise.
For those looking for an independent and updated view of current MariaDB performance, I encourage you to read a recent blog by Steve Shaw and see how our latest engine compares when pushed to its actual limits. You can find more details at Scaling Databases in the Chiplet Era: Lessons from MariaDB Performance Engineering, which provides a window into the current state of MariaDB performance.
Looking Ahead
My first year at MariaDB has been defined by speed – both in our database engine and our pace of innovation. We are committed to transparency and performance engineering that translates to real-world value for our users. I couldn’t be prouder of the team, and we’re just getting started.
Stay tuned – there is plenty more to come in 2026.
