We're announcing a technical partnership with Tower that brings together real-time analytics and Python-native data orchestration. For data teams building AI-powered systems, this partnership delivers a unified architecture that eliminates the traditional trade-off between high-speed analytical serving and governed historical data.
Who Is Tower?
Tower provides teams with data infrastructure as a service to build, schedule, and monitor complex data pipelines using the tools they already know. Tower handles the operational complexity of DataOps so teams can focus entirely on building resilient, mission-critical workflows without the infrastructure overhead. Ship product, not infrastructure.
Brief Recap on VeloDB
We built VeloDB to deliver sub-second query performance at scale, no matter how fast your data changes or how many users and AI agents are querying at once. VeloDB is a real-time OLAP database that handles analytical queries, full-text search, and vector-based AI retrieval in a single system. Whether your team runs customer-facing dashboards, log analytics, or RAG pipelines for context engineering, VeloDB consolidates what previously required three or four separate databases into one serving layer.
Speed alone is not enough. VeloDB maintains sub-second latency at thousands of concurrent queries, even under heavy data updates and complex multi-table joins. As agentic AI systems multiply query volumes by orders of magnitude, that concurrency ceiling matters more than ever.
VeloDB and Tower: Speed Meets Orchestration
From a technical standpoint, VeloDB and Tower unify the historical and real-time tracks of your data architecture.
VeloDB acts as the high-speed analytical engine and serving layer. It ingests data streams directly for sub-second performance and, because we natively support Iceberg (V3) and other open formats, VeloDB can also read and join data residing in your lakehouse directly. This lets teams consolidate their entire serving layer: ad hoc queries over historical data and real-time workloads run in the same place, against the same system.
Tower sits above your data estate as the intelligent orchestration layer. On the historical side, Tower manages complex pipelines feeding data from Snowflake, Databricks, and other platforms into your Iceberg lakehouse. On the real-time side, Tower directs data streams and change data capture into VeloDB for immediate processing. Critically, Tower also runs the jobs that extract fresh data from VeloDB and merge it back into the Iceberg lakehouse, keeping your historical archive current without brittle custom pipelines.
The result: developers get a unified control plane with full visibility and governance over data movement across the entire hybrid stack. Applications and end users get an ultra-fast, high-concurrency serving layer through VeloDB.
Unifying Your Serving Layer: Bridging Batch and Real-Time
Data engineers today routinely string together disjointed tools, manage brittle dependencies, and maintain entirely separate architectures for batch and streaming data. Add AI agents and large language models to the mix, and the complexity compounds fast.
The core challenge is building a serving layer that combines fresh real-time insights with massive historical context. Fragmented stacks create data inconsistencies, slow time-to-insight, and growing operational overhead. As AI agents multiply query volumes, the cost equation shifts dramatically. Organizations can no longer afford to pay for idle capacity or architect separate systems for every new workload.
VeloDB and Tower address this with a modified lambda architecture that simplifies DataOps while radically boosting performance. Tower orchestrates the full data lifecycle: it manages pipelines into an open, governed Iceberg lakehouse and directs real-time streams into VeloDB. Tower then acts as the data-bridging mechanism, extracting the newest data from VeloDB and merging it seamlessly into the lakehouse so your historical archive stays in sync.
This makes VeloDB the single analytical engine for BI, dashboards, and AI applications. Because VeloDB maintains sub-second latency at thousands of concurrent queries, AI agents and human users share the same system without competing for resources. Pipelines run continuously, orchestrated by Tower. VeloDB serves sub-second responses for dashboards, embedded analytics, and agentic context retrieval, giving your users a real-time pulse on your entire business.
You get the infinite scale and robust governance of an Iceberg-based data lake combined with the rapid interactivity of a world-class real-time analytical engine. Critically, you get them in one architecture, without the brittle glue code and reconciliation bottlenecks that come from stitching together separate systems.
This partnership simplifies the modern data stack for teams building at the intersection of real-time analytics and AI. By bringing Python-native orchestration and real-time analytics together, VeloDB and Tower help data teams stop wrestling with infrastructure and start delivering the AI-driven data systems their products demand.
Try Tower.dev and VeloDB Cloud today
Join the Apache Doris Community and Tower.dev community to learn more
