Gabriel Ciciliani will be speaking at Percona Live 2026 about #MySQL internal temporary tables and the hidden behavior that shapes query performance.
If you’re coming to Percona Live, join the session and come say hi 👋 to Gabriel and the Releem team.
See you at #PerconaLive 🙂
We finally moved Releem platform from Azure back to Google Cloud.
Not because migration is fun. It is not)
Google now covers our cloud costs through its startup program, so moving back made sense.
Our infrastructure history is a small startup story by itself:
- We started on AWS.
- Two years later, we moved to GCP.
- We did not get into Google’s startup program then, so we moved to Azure.
- Now we are back on GCP.
Each time, we moved without service interruption. And each time, Releem became a little less tied to one cloud.
The hardest part is always the databases. Moving terabytes is not a “copy files and restart” service. You have to think about replication, consistency, failover, and all the boring details that decide whether users notice anything.
This migration also pushed us to do something we should have done earlier:
1. More Terraform and installation scripts.
2. Less hidden, cloud-specific knowledge.
Most importantly, we removed the last AWS Lambda dependency ( yeah, old services still used it )
The result is that Releem can now run the same way across clouds, or on dedicated servers.
Now back to building)
We simplified how recommendations are shown in Releem SQL Query Analytics.
You can now see analysis and recommendations for top and slowest queries without extra clicks.
If a query is fine, you’ll still see that it was checked and when it was last analyzed.
🚀 What’s new at Releem in March?
🔍 Recommendations for all queries are now visible directly in Query Analytics
🧠 Custom SQL query optimization
🖥️ WHM/cPanel + WHMCS integration
☁️ AWS Aurora Serverless support
📊 Deadlocks can now be exported to CSV
Blog post👇
We shipped optimization for custom SQL queries.
Developers can paste any SQL query and get recommendations before it goes to production.
Releem automatically checks the query execution plan, schema, and database metrics to provide suggestions.
In March, we’ll be at CloudFest in Germany (23–26 March), and you can find us in Startup Alley.
If you’re planning to attend, let us know so we can connect in person and discuss database performance, automation, and partnership opportunities.
We shipped optimization for custom SQL queries.
Developers can paste any SQL query and get recommendations before it goes to production.
One customer told us they built an AI agent that checks every PR.
Now they want MCP access to Releem so the agent can request recommendations automatically before merge.
It makes sense. If developers are working with agents, the product should work with agents too)
Last day in Brussels, I had a chance to visit the Summit for MySQL Community organized by Percona.
Great organization, lots of interesting discussions and meetings.
I met with Ewen, founder of SIXTA. We had a good conversation and realized we’re both working on a similar idea: a Database Advisor that works on top of existing monitoring data. But we’re solving different problems.
At Releem, we focus on analysis, recommendations, and safe automation for day-to-day DBA tasks. Ewen is working on automating root cause analysis when issues arise.
Photo at the end with friends. Sadly, many had already left by then.
Heading to FOSDEM this weekend.
I’ll also attend MariaDB Days, the MySQL Community Summit, and MySQL Days.
I’ll be speaking at MariaDB Day about what we’ve learned working with teams running MariaDB without a dedicated DBA.
I’ll share how teams are using Releem as a database advisor to reduce manual work and make database operations more predictable.
Unfortunately, I missed my MySQL talk at MySQL Days today due to visa guesswork, but looking forward to catching up with everyone in Brussels!
I’ll be speaking at the AI DBA conference about how teams without DBAs manage database servers using Releem.
I’ll share how teams manage databases today, where they struggle, and what we learned while building a database advisor to address these problems.
I’ll also walk through how Releem works in practice and explain the architecture behind it.
Looking forward to the discussion.
The registration link 👇
I’d like to share my experience with DMARC report analysis using AI.
A few months ago, I set up a MySQL database to store DMARC email reports for Releem. I used a simple schema from a public n8n workflow.
Last week, I needed to analyze this data. Instead of spending time writing SQL queries and doing everything manually, I decided to try Claude with the MySQL MCP Server ( https://t.co/yO8fK8oY5y )
The setup was straightforward. I connected it to Claude using Docker in about 5–10 minutes.
Then I asked Claude to analyze DMARC data from the last three months. It automatically queried the database and pulled the relevant data.
Within a few minutes, I had a complete report with recommendations on what to do next 🤯
Thanks to @ask_dba , great work 👍
2025 was a great year, both personally and for the Releem team. I met many interesting people and learned a lot of new things.
It started with my first visit to FOSDEM, where I got to meet the amazing MySQL and MariaDB communities, including Monty, the creator of both.
I had a 1-month travel adventure with my family. We visited 3 different cities.
Later in the year, I summited Mount Kazbek 🏔️ and ran my first marathon 🏃 Two very different challenges, but both equally memorable.
👨💻 Releem Team
- My co-founders Victor and Dmitry started working full-time, boosting our development and release cadence
- The TuningWizard team and their founder, Gabriel, joined Releem, bringing production-grade SQL query optimization experience in-house.
- We gave engineering talks at Percona University and Big Data Conf 2025
- Ran 100+ live demos and handled 1400+ support questions
- Shipped 1,292 commits and did ~200 platform deployments
- Launched ~50 marketing experiments. Most of them failed 🙃
🚀 Releem Product Highlights
We shipped a lot this year, focusing on simplifying and automating database management operations:
- Real-time deadlock detection with root cause analysis
- Live process list for long-running/blocking queries
- Smart schema checks including missing, unused, and redundant indexes detection.
- Full config history with before/after performance metrics
- Full support for GCP Cloud SQL
- Per-server access control, user roles, and... yes, Dark Mode 😎
📊 2025 Releem in Numbers
- 348B SQL queries collected 🤯
- 808M API requests to Releem Platform
- 714k schema recommendations
- 6M SQL query recommendations
- 571,000 deadlocks detected and analyzed
- 53,000 config tuning suggestions
- ~1,000 new users joined
Huge thanks to our users, early adopters, and the database community for the support 🙌
We're learning fast, shipping faster, and not stopping in 2026 🦾
Happy New Year! 🎄
🚀 Product update from Releem
This month we shipped:
✅ User roles and dashboard sharing
🌙 Dark mode
🧠 Smarter buffer pool recommendations for MySQL & MariaDB
⚙️ Initial configuration for new MySQL installations
Read the full update 👇
We finally shipped dark mode to Releem Dashboard 🎉
It took longer than expected, lots of tiny style fixes.
We quietly rolled it out, and a few dozen users enabled it right away.
Today I saw an interesting poll from Peter Zaitsev:
“Do you think you can do better than AI tuning your database configuration?”
It reminded me of how Releem got started.
A few years ago, I was working at a small consulting company. We managed hundreds of MySQL servers and manually tuned configurations as needed.
The problem I faced: Junior engineers/DBAs didn’t have enough experience to safely tune configs, and manual tuning was time-consuming and error-prone.
I wrote documentation to help, but docs alone didn’t prevent mistakes.
So, I built a bash script and a small cloud service to generate MySQL / MariaDB configurations based on workload, system resources, and dbms version. That early version was completely rule-based.
I quietly rolled it out to the team. After a week, a senior teammate said:
"Roman, this script will never tune better than me."
Fair enough.☝️
About 8 months later, I asked him:
"When was the last time you tuned a config manually?"
His answer:
"I haven’t touched anything for a few months. The tool just works." 👀
That tool ended up saving our team hundreds of hours.
That's when I realized it could be valuable for others too.
I started collecting a waitlist.
That script became the MVP, and that was the beginning of Releem.
As for Peter’s poll… maybe the next question should be:
"Do you have time to tune database configuration better than AI?"