Blueprinting the Cloud: Unifying and Automatically Optimizing Cloud Data Infrastructures with BRAD -- Extended Version

TL;DR

BRAD introduces blueprint planning, a cost-based optimization framework that virtualizes cloud data infrastructures by unifying engine selection, data placement, provisioning, and query routing across multiple cloud database services. It uses a graph neural network trained on logical SQL features to predict query run times and data scanned, augmented with analytical models to adjust for provisioning and load, enabling greedy beam-based search over a space of blueprints. The system serves queries through a single SQL interface while automatically achieving performance targets and minimizing operating costs, with transitions between blueprints accounted for in optimization. Across a diverse QuickFlix workload, BRAD delivers substantial cost savings (up to 13x) relative to serverless autoscaling or HTAP baselines while maintaining latency constraints, demonstrating practical scalability and adaptability to workload shifts.

Abstract

Modern organizations manage their data with a wide variety of specialized cloud database engines (e.g., Aurora, BigQuery, etc.). However, designing and managing such infrastructures is hard. Developers must consider many possible designs with non-obvious performance consequences; moreover, current software abstractions tightly couple applications to specific systems (e.g., with engine-specific clients), making it difficult to change after initial deployment. A better solution would virtualize cloud data management, allowing developers to declaratively specify their workload requirements and rely on automated solutions to design and manage the physical realization. In this paper, we present a technique called blueprint planning that achieves this vision. The key idea is to project data infrastructure design decisions into a unified design space (blueprints). We then systematically search over candidate blueprints using cost-based optimization, leveraging learned models to predict the utility of a blueprint on the workload. We use this technique to build BRAD, the first cloud data virtualization system. BRAD users issue queries to a single SQL interface that can be backed by multiple cloud database services. BRAD automatically selects the most suitable engine for each query, provisions and manages resources to minimize costs, and evolves the infrastructure to adapt to workload shifts. Our evaluation shows that BRAD meet user-defined performance targets and improve cost-savings by 1.6-13x compared to serverless auto-scaling or HTAP systems.

Figures (15)

• Figure 1: Query performance and operating costs of the same workload on two data infrastructure designs.
• Figure 2: BRAD's blueprint planning life cycle.
• Figure 3: A detailed view of BRAD's architecture and its end-to-end blueprint planning life cycle.
• Figure 4: The query featurization used by our model, which predicts a query's run time and the amount of data it scans.
• Figure 5: BRAD reduces cost while maintaining p90 latency constraints (shaded region). The dotted (solid) vertical lines indicate when a new blueprint is chosen (takes effect).