What Lakehouse//RT and LTAP Are—and Why They Matter
Databricks Lakehouse//RT and LTAP are complementary data lakehouse innovations that unite transactional and analytical workloads on the same governed storage, enabling real-time analytics and operational processing without copies, ETL pipelines, or separate serving systems. For decades, enterprises ran transactional systems for operations and analytical systems for reporting, connecting them with fragile change data capture pipelines and complex replication. Databricks is attacking this split at its root. Lakehouse//RT brings millisecond real-time analytics directly into the lakehouse architecture, while LTAP (Lake Transactional/Analytical Processing) unifies OLTP-style transactions and OLAP analytics on one copy of data in the lake. Together, they aim to remove the old trade-off between fresh operational data and scalable analytics. Instead of maintaining dual stacks and reconciling copies, teams can query governed Delta Lake and Apache Iceberg tables as their single source of truth for both transactional and analytical work.
Lakehouse//RT: Real-Time Analytics Built into the Data Lakehouse
Lakehouse//RT is Databricks’ new real-time analytics layer, designed to bring low-latency querying directly to the data lakehouse architecture. It runs on Reyden, a compute engine built for high concurrency and millisecond response times on governed Delta Lake and Apache Iceberg tables. According to Databricks, Lakehouse//RT achieved sub-100 millisecond latency at 12,000 queries per second and delivered up to 16x better performance than traditional real-time serving stacks in customer tests. Because it queries the canonical tables in place, it removes the need for separate serving databases, proprietary formats, and sync or CDC pipelines. Every Lakehouse//RT query is governed by Unity Catalog, so there is no extra permissions layer to manage. The result is real-time analytics on fresh operational data without building and operating a parallel stack for low-latency serving.
LTAP: A Unified Architecture for Transactional and Analytical Workloads
LTAP (Lake Transactional/Analytical Processing) is Databricks’ new architecture that unifies OLTP-style transactional workloads with OLAP analytics on a single copy of data in the lake. Instead of pushing both workloads into one monolithic engine, LTAP unifies them at the storage layer. Transactional data written through Lakebase—Databricks’ Postgres-compatible engine on open object storage—is immediately queryable for analytics in the lakehouse without ETL, replicas, or hidden CDC pipelines. Databricks describes this as the first LTAP platform, combining Lakebase and the Lakehouse under one governance model and storage layer for operational, analytical, and streaming data. Lakebase already serves thousands of customers and handles 12 million database launches per day, showing that the transactional side can scale independently. Analytical and transactional workloads remain isolated in compute, but share the same open storage, allowing each to scale without impacting the other.
How LTAP and Lakehouse//RT Eliminate Dual Systems and Latency
Together, LTAP and Lakehouse//RT target the long-standing separation between operational databases and data warehouses. LTAP ensures that all transactional and operational writes land in open object storage, in the same data lakehouse that powers analytics. Lakehouse//RT then brings real-time analytics to that shared storage, so applications, dashboards, and AI agents can query fresh data with millisecond-level response times. This combination means teams no longer need to maintain separate OLTP systems plus analytical warehouses plus specialized serving layers, each with its own data copies and governance model. Instead of building and monitoring CDC jobs and synchronization pipelines, they query a single governed data lakehouse that already contains both transactional and analytical data. For organizations building AI applications and agents that must read, reason, and act in near real time, this unified architecture provides a path to scale without the usual complexity and data staleness.
Implications for Real-Time AI and the Future of Lakehouse Architecture
The most immediate impact of combining Lakehouse//RT with LTAP is on AI-driven and agentic applications that depend on real-time analytics. These agents run continuous loops of reading data, reasoning over it, and acting, and they need both transactional context and analytical aggregates. With LTAP, the operational state of applications lands directly in the data lakehouse; with Lakehouse//RT, that same state is queryable at sub-100ms latency for tens of thousands of concurrent users and agents. This tight integration reduces architectural sprawl and allows teams to focus on application logic instead of data plumbing. It also reinforces the data lakehouse as the central platform for all workloads—streaming, transactional, and analytical—rather than a warehouse add-on. As more organizations move to this kind of unified lakehouse architecture, separate operational databases, warehouses, and serving clusters are likely to become optional instead of mandatory building blocks.
