Conference50min
10 Billion Records vs. The Regulator: An Architectural Post-Mortem of an 11TB Exadata Failure
A major bank incident triggered a BaFin investigation after an 11TB Oracle Exadata system failed at 10 billion records. This session examines emergency mitigation, a missed query-window check, cache warming, shifting deduplication to writes, and designing zero-loss dual-write migration to Cloud PostgreSQL.
talk.summaryAiDisclaimer
Martin LadeckyCommerzbank
talkDetail.whenAndWhere
Friday, October 9, 10:40-11:30
TBA 3
talks.roomOccupancytalks.noOccupancyInfo
What happens when your architecture doesn't just fail a performance test, but triggers a national regulatory investigation? In this session, we go inside a critical system failure at a major bank that led to a formal report to BaFin and a complete rethink of our data strategy.
Dealing with 11TB of data on Oracle Exadata taught us that at a certain scale—specifically, the 10-billion-record mark—standard indexing strategies and read-side duplicate checks don't just slow down; they collapse. We will reverse-engineer this incident by walking through the five critical architectural milestones:
Dealing with 11TB of data on Oracle Exadata taught us that at a certain scale—specifically, the 10-billion-record mark—standard indexing strategies and read-side duplicate checks don't just slow down; they collapse. We will reverse-engineer this incident by walking through the five critical architectural milestones:
- First Aid: Immediate application-layer mitigation steps to halt cascading timeouts.
- The Refactoring Oversight: How a missing 10-day query window check, removed years prior, silently waited to trigger an 11TB full scan.
- The Cache Bottleneck: The high-stakes "do nothing" war room decision to wait for the Exadata Smart Flash Cache to warm up and recover.
- The Structural Flip: Moving 10kB of complex analytic queries from read-side deduplication to a write-side integrity layer.
- The Exit Strategy: Designing and planning a zero-loss dual-write ingestion architecture to migrate 10 billion records to Cloud PostgreSQL without CDC.
Martin Ladecky
Martin Ladecký is an enterprise architect and systems engineer dedicated to building resilient, high-scale data infrastructure and promoting clean engineering practices. Operating within heavily regulated financial environments, he specialises in optimising core Java workloads, container stability, and complex multi-terabyte database ecosystems.
As an active contributor to the European open-source and tech community, Martin balances his corporate work at scale with meaningful "Tech for Good" initiatives. He organises the Czech Java User Group and contributes to Česko.Digital to help NPOs change lives through technology.
When he isn’t diving into thread dumps or troubleshooting connection pools, you can usually find him exploring local running trails or enjoying a great espresso.
As an active contributor to the European open-source and tech community, Martin balances his corporate work at scale with meaningful "Tech for Good" initiatives. He organises the Czech Java User Group and contributes to Česko.Digital to help NPOs change lives through technology.
When he isn’t diving into thread dumps or troubleshooting connection pools, you can usually find him exploring local running trails or enjoying a great espresso.