[IT Daily] “More than 60% of DB performance failures start with SQL. You must optimize SQL before expanding hardware.That is the most realistic cost-reduction strategy a CIO can implement right now. We call this 'Tune First, Expand Later.'”

Managing Director Jang Cheol-woong of the Solution Business Division at Openmade Consulting presented this solution for infrastructure cost issues in the AI era during a recent interview with this journal (IT Daily/Computer World). Based on years of field experience in the IT industry ranging from hardware to System Integration (SI), Big Data, and DB, Managing Director Jang is currently overseeing sales for ‘Query Medic,’ an AI-based SQL auto-optimization solution at Openmade Consulting.

“Era of Infrastructure Inflation... No Financial Control Without SQL Optimization”

Corporate IT environments this year are in a complex situation. As the global AI server market grows to approximately $220 billion, prices for GPUs and memory semiconductors are skyrocketing, and DRAM prices are expected to rise by more than 50% year-on-year.

Consequently, a phase of ‘Infrastructure Inflation’ is ongoing, where securing stock for server expansion is difficult, and even if secured, costs increase exponentially.

The situation is not much different for companies that chose Cloud migration for cost reduction. The Cloud environment is structured to charge fees based on resource usage. Migrating non-optimized SQL directly leads to cost increases due to unnecessary I/O and CPU consumption.

Managing Director Jang diagnosed, “No matter how much you expand infrastructure, if the SQL running on it is inefficient, costs will leak endlessly. From a CIO's perspective, whether buying servers or using Cloud, financial control is becoming impossible without SQL optimization.”

“SQL Tuning Limit of 3-4 Cases per Day... Creating Management Blind Spots”

The issue is who performs this SQL optimization. According to Managing Director Jang, field DB management personnel are too busy handling operational tasks to afford performance improvements. A skilled DBA can only handle a limit of 3-4 SQL tuning cases per day.

Managing Director Jang said, “Lists of problematic queries are managed by individuals or teams based on experience. While known issues are managed, it is truly dangerous when unknown ambushes accumulate and explode as failures all at once one day.”

These issues are also acting as bottlenecks in development sites. For instance, if an IT organization has 300 developers, each writes SQL, but there is a shortage of tuning experts to verify the performance of the queries they've written. Even if a developer wants to request tuning, there is no one to accept it, and DBAs are too busy with their own operational tasks to look at others' queries.

Managing Director Jang pointed out, "Both the development department and the DB operation department desperately feel the need for tuning, but they are in a situation where they cannot solve it for each other," adding, "That is the biggest structural problem currently facing domestic IT organizations."

“탐지부터 튜닝·검증까지 전 과정 AI로 자동화”

Query Medic was born to solve these problems. Openmade Consulting has accumulated field experience in DB performance management by dispatching DBAs, tuners, and modelers primarily to the financial sector for 21 years, and integrated this know-how to develop and operate ‘OpenPOP.’ OpenPOP is a solution that checks, verifies, and controls performance throughout the entire process from SQL development to operation, and preemptively prevents failures by applying a 24/7 failure prevention process in operational environments.

According to Managing Director Jang, there were continuous requests from clients asking, "Can't you automate the tuning itself instead of just providing a guide?" While OpenPOP could identify the problems, the actual task of fixing the SQL still fell to humans.

Consequently, full-scale Query Medic R&D began two years ago when AI technology was rapidly advancing, combining LLM reasoning capabilities with OpenPOP's detection and verification engine. The core is automating the entire process, from automatic extraction of underperforming SQL to generating optimal tuning results and verifying them. It was officially launched in October last year.

Detailed tuning logic derived from field experience was also reflected. In actual DB operational environments, the optimal tuning method often varies for the same SQL depending on when and for what purpose it is executed.

For example, for OLTP queries that must respond immediately to user requests during the day, the key is returning results even 0.1 seconds faster. However, for Batch queries that process large amounts of data for hours at dawn, it is more important to finish the task within a set time by reducing the total throughput rather than individual response speed. Query Medic automatically identifies the type and execution pattern of the target SQL to apply optimization logic for each situation.

Managing Director Jang explained, “If you put a query into general AI, you get a textbook answer, but that often doesn't work in the field. The quality of the results is different because the engine contains years of experience gained in the financial sector, such as judgment criteria on why a certain tuning method is actually better in a specific situation.”

“The Key is Whether the SQL Fixed by AI is Reliable”

Managing Director Jang cited the 'AI-based Verification' function as Query Medic's key differentiator. He explained that while one can obtain SQL tuning results with general-purpose AI tools, whether those results can be applied directly to practice is a completely different matter.

Managing Director Jang said, “If you put SQL into ChatGPT or Gemini, a plausible tuning plan comes out. However, since the consistency of the results is not guaranteed, it cannot be applied to operational environments. In fact, many companies have tried SQL tuning by introducing their own LLM models, but most gave up on practical application because they couldn't trust the results.”

Query Medic directly executes the AI-generated tuning plan in a DB environment containing actual operational data and automatically compares and verifies if the results and record counts before and after tuning match exactly.

Managing Director Jang emphasized, “There is no solution yet, domestically or globally, that automates the full step from tuning to verification. This verification function is truly special.”

“One DBA Does 60 Cases a Month, Query Medic Does Over 1,000 Cases a Day”

According to Managing Director Jang, a single DBA can handle approximately 3 SQL tuning cases per day. Simple math shows they handle 60 cases per month (based on 20 working days) and about 720 cases per year. Query Medic operates automatically 24/365 and can tune more than 1,000 SQLs per day. Furthermore, it preemptively eliminates potential failure SQLs that humans might not yet recognize by performing a full inspection, offering failure prevention benefits.

Managing Director Jang explained, “What humans manage is only a portion of what is recorded or known empirically by the person in charge. Because Query Medic performs a full inspection of all SQL, it can catch 'ambushes' in advance that no one knew about until they suddenly exploded as a failure one day.”

He stated that expectations in the field for Query Medic are high. Managing Director Jang said, “When we introduce Query Medic, the reactions from developers, DB personnel, and CIOs are the most enthusiastic. From a developer's perspective, they gain a means to immediately verify the quality of the SQL they write. From a CIO's perspective, they secure a tool to boost operational efficiency and control costs without expanding servers or hiring more staff.”

He added, "One financial client tried to achieve business results by introducing various AI models for 5-6 years but did not get satisfactory results. After a Query Medic demo and POC, the IT department evaluated that 'Query Medic is the only solution that can produce actual results with AI.'"

Currently, the adoption of Query Medic is expanding, primarily in public institutions and the financial sector. There has been a case of supplying Query Medic for a system reconstruction project in a public institution, and numerous financial companies and public institutions are conducting POCs or reviewing adoption.

Interest in the manufacturing sector has also been increasing recently. Managing Director Jang said, “Major domestic manufacturing companies frequently experience performance failures due to the quality issues of SQL written by local developers while operating overseas production lines. Since production line stoppages result in massive cost losses, inquiries from the manufacturing sector have significantly increased as the awareness spreads that a significant part of the cause lies in the SQL running on the DB.”

“I Will Plant the Principle of ‘Tuning First’ in the Field”

Query Medic currently supports Oracle environments and is rapidly expanding its supported DBs. Development for PostgreSQL is nearly complete, and support for Tibero, a domestic DB, is scheduled within the second quarter. Development for MySQL is underway with a goal of the second half of the year. From 2027, there are plans to provide a SaaS-based version.

Managing Director Jang stated, “This is possible because of the technical trust accumulated over 21 years in the specialized field of DB tuning. I want to plant the belief in the market that systems with Query Medic are safe.”

He added, “The era where DBAs stay up all night examining queries one by one must end. I will work in the field until the principle of ‘Tune First, Expand Later’ becomes common sense for CIOs.”

Sales inquiries for product implementation and consulting

02-6310-6167 / qm@openmade.co.kr 

Source : IT Daily (http://www.itdaily.kr/news/articleView.html?idxno=238877)

Reporter seungyang@itdaily.kr

#SQLTuning #AITuning #DatabasePerformance #QueryMedic #OpenmadeConsulting #DBOptimization #AutomationSolution #2025NewProduct #DBA #Developer #ITOperations #CostReduction #PerformanceImprovement #AIAutomation #DBPerformanceManagement #InfrastructureCostReduction #CIO