We solve Microsoft SQL Server performance problems, plan and build highly available systems, and support enterprises, MSPs and data center operators, and software vendors in day-to-day operations.
Performance engineering and infrastructure expertise in one team. With our own tools, local and traceable.
For many years we have been solving performance problems in SQL Server applications: in the interplay of application, queries, plans and platform. With experience from many environments and our own tools, which deliver evidence instead of guesswork.
More on SQL Server performance →
The system is slow, and time is short. We investigate at short notice, narrow down the cause and agree on immediate measures with you.
Get help quickly → TargetedHow well is the system set up? Why has it behaved differently since the update? We investigate in a structured way and deliver prioritized recommendations.
To the SQL Server Health Check → OngoingCritical applications continuously in view: observing, investigating, improving and verifying the effect, within the agreed scope.
Continuous support →Your application runs on SQL Server and has to stay available and fast. We solve performance problems, plan the architecture and build it with you.
You are responsible for operations for your customers. We complement your team where SQL Server demands particular depth: architecture, difficult cases and planned changes.
The same problem shows up at customer A and months later at customer B. With QX Fleet, development and support see patterns across all installations.
“PSG has been helping us very successfully since early 2020 and enables our software developers to focus more on their creative tasks again. […] The technical level of the consulting impresses us again and again.”
“Instead of generic best practices and piecemeal symptom fixes, they understood our development approach and the nature of our application systems and, starting from there, worked out concrete optimizations.”
“What impresses me about working with PSG, besides the exceptional, field-proven expertise, is above all the ability to get support at short notice.”
Five layers that build on one another. You engage the ones you need. We use our tools throughout so that every finding is verifiable.
We plan and implement SQL Server infrastructures. We align architecture, edition and core count with availability, performance and cost-effectiveness. Because we don't sell licenses, we advise without a sales interest of our own.
Ongoing SQL Server administration, owned jointly with your team, with a regular look at the state of your systems.
Our core: solving performance problems, urgently or in a targeted way, and continuously improving critical applications with PSG MX. Month by month as MX-Care, or in intensive phases as MX-Intensive.
Evidence instead of guesswork. QX analyzes the execution history: which query, which plan, since when, and what actually changed after a modification. The Recorder captures the evidence of sporadic problems before it disappears.
Performance knowledge across all installations of the same software. We show software vendors whether an anomaly affects the product, a group of environments or a single installation, and how a new release behaves in production.
The performance timeline shows CPU, wait times and executions over time and, below that, which plan of which query caused the load.
Executions, CPU, reads and duration appear right at the line in the procedure's source code, along with the number of distinct plans.
Every finding names its evidence, what is unknown and the next step to check. Here: a risk from cached statistics of a temporary table.
Prefer to talk right away?
Every engagement follows the same pattern, whether a one-time analysis or ongoing support.
What exactly should get better, and how would you notice?
Collect the data that can actually answer that question.
State the finding, evidence, hypothesis and limits openly.
With operations, development and, where applicable, the vendor.
Compare before and after instead of just guessing.
We have been using data mining for SQL Server analysis for years, long before generative AI and LLMs. Our methods run locally and deterministically: with the same inputs and method versions, every result can be checked again.
observation A central query has been switching between two execution plans since a cutoff date. evidence Query Store history: plan changes, run times, affected time windows. hypothesis Parameter-dependent plan after a statistics update. uncertainty Influence of parallel maintenance jobs not yet ruled out. measure Agreed change in a test window. verification Compare the same metrics before and after the change.
The analysis runs without calling external AI services. Which data flows where, we state specifically for every engagement.
Same inputs, same method version, same result. That keeps every finding repeatable.
Every statement names its evidence, its uncertainty and the way to verify it.
PSG is a consultancy from Hamburg specializing in Microsoft SQL Server. Because we started in 1994 as a software house, we approach every topic as developers: with code, automation and our own tools. These are our areas of expertise, each with what the statement rests on.
Finding causes in queries, plans, wait times and configuration, urgently or in a targeted way.
SQL Server projects since 1998Tracing plan changes, regressions and their causes over time.
our own tool: PSG QXAlways On Availability Groups, failover clusters, disaster recovery, edition and cores.
own infrastructure team, no license salesOperations, standardization and maintenance of many servers, generated and automated instead of by hand.
Projects with more than 1,000 serversAssessment, migration and operation of memory-optimized tables.
Migrations since 2015Speeding up analytical queries on large data volumes.
Microsoft Partner Conference 2012Considering performance in development and architecture from the start.
Trainings and tools since 2016Recognizing patterns across all customer installations of the same software.
QX Fleet for software vendorsTraining, coaching and workshops so that your team understands the connections.
How development, project management and product owners consider performance from the start instead of repairing it later in operations.
Performance Troubleshooting & Support: how SQL Server works internally and how to narrow down bottlenecks systematically.
Designing data access with the repository pattern and clear abstraction so that complexity stays manageable.
Custom formats built around your questions and systems, from a single execution plan to an architecture decision.
Availability, performance, health check, potential analysis, data warehouse audit or migration: compact, with clear next steps.
We work remotely, through the channels your IT specifies: at short notice and flexibly, even when only an hour on the systems is needed. Same time zone, and we work in German or English, whichever suits your team. And if it really has to be on site, we board a plane.
From Hamburg across Germany, for enterprises, operators and software vendors.
We support complex and critical SQL Server environments in Austria.
SQL Server in Austria →We support complex and critical SQL Server environments in Switzerland.
SQL Server in Switzerland →PSG was founded in Hamburg in 1994 as a software house. We have been working as consultants for Microsoft SQL Server since 1998. We come primarily from software development, not from administration. That shapes our work to this day: we read the code behind the query, automate what repeats, and build our tools ourselves.
Today, performance engineers and infrastructure colleagues work together in one team: the former investigate what is happening in your systems, the latter plan and build highly available platforms. On the team you will find certifications, talks at professional conferences and a former Microsoft MVP (2014–2016). The PSG Stack is an offering of PSG Projekt Service GmbH.
Think you would fit our team? Write to us at jobs@psg.de.
Early with new SQL Server technologies, on big stages and in demanding projects, and with technical articles and talks in the national and international community.
Among other things, table compression, at the time a brand-new SQL Server 2008 feature. Row and page compression save storage space and often I/O as well, because fewer pages have to be read for the same data.
Live on the big Microsoft stage, in front of the trade fair audience.
Repository Driven Design: structures and logic are described once, declaratively, in a central repository. From that, whatever each target environment needs is generated, instead of maintaining scripts for each server individually.
1,000+servers in the first project, today also offered as training
New with SQL Server 2012: data is stored column by column and highly compressed, and queries process whole blocks of values at once. Analyses over large tables often become many times faster as a result.
Our message to the partners: this makes customers very happy.
Since SQL Server 2014, tables can reside entirely in memory and procedures can run natively compiled, without locks and latches. With many parallel transactions, this can be a big leap.
Our principle: the load test with your real workload decides, not the demo.
Performance Driven Development: performance belongs in development and architecture from the start, instead of being repaired later in operations. With continuous measurement, jointly understood expectations, and with project management and product owners at the table.
Instead of a single person, a well-coordinated team looks after your SQL Servers: in monthly packages from Bronze to Gold, with regular checks and reachable on business days. MX was added, our monitoring framework, which measures in your environment.
Today's tools MX AIOps and PSG QX come from this line.
Monitoring that does not only report what has already happened: from the measurement history we read trends and recurring patterns, such as data growth or gradual degradation. That lets you plan maintenance and capacity before things get tight.
Local and deterministic: data mining, long before generative AI.
In the first conversation we clarify whether and how we can help. We reply within 24 hours on business days.
We arrange license keys, renewals and trial access directly with you.
