Beyond the Forklift: Unlocking Revenue-Critical Workloads on Iceberg

Apache Iceberg™ has become the open standard and leading open table format for modern data platforms, yet most adoptions approach migration as a forklift. Governance improves, storage is standardized, and BI workloads run reliably; but the most demanding analytics, the ones closest to revenue and customer experience, are rarely considered in scope for Iceberg.

Those SLA-bound data products—embedded dashboards, fintech merchant cash flow views, surge pricing, fraud detection, incident response—don’t simply go away when Iceberg isn’t built to serve them. If the platform isn’t engineered for deterministic p95/p99 latency, high concurrency, and fresh data on Iceberg tables, the requirement resurfaces elsewhere. What was intended to be a shared system of record becomes relegated to a feeder layer, pushing data into downstream systems where it is re-stored, re-governed, and re-queried outside of Iceberg.

This session outlines a different model: Apache Iceberg™ as the open system of record, paired with Apache Pinot™ as a purpose-built execution layer for SLA-driven analytics that enforces deterministic SLAs directly on Iceberg tables, eliminating shadow stacks and restoring architectural coherence for revenue-critical insights.