Summary

Flashpoint is a global leader in threat intelligence, providing organizations with actionable insights to mitigate cyber risks and protect critical assets. They rely on Looker and BigQuery to power complex business intelligence dashboards and internal analytics that drive their mission-critical services. By partnering with Alvin, Flashpoint successfully addressed escalating on-demand costs while simultaneously improving query performance through autonomous optimization and intelligent caching.

‍

The Challenge

Flashpoint's Looker environment represented their primary source of BigQuery spend, operating historically on an on-demand pricing model. Because on-demand billing is based on bytes scanned without an upper limit, even a single query could lead to unpredictable cost spikes. The engineering team faced a persistent dilemma: transitioning to capacity-based pricing risked performance degradation during peak hours or paying for underutilized slots during lulls. Without a way to treat queries individually, they were forced to balance financial predictability against the rapid results required for their time-sensitive BI workloads.

‍

The Solution

To eliminate these trade-offs, Flashpoint integrated Alvin’s autonomous agent as a proxy service between Looker and the BigQuery API. The setup required zero code changes and enabled autonomous optimization of their Looker workloads.

• Dynamic Query Routing: Alvin fingerprints every Looker query in real-time, analyzing historical slot usage and bytes processed to route each request to the most cost-effective billing model—either on-demand or capacity.
• Query Rewriting for Cache Hits: Alvin parses SQL into an Abstract Syntax Tree (AST) to inline expressions common in Looker that break the BigQuery cache. This allows a significantly higher percentage of queries to be served from the cache at zero cost.
• Autonomous Reservation Management: To protect the performance of time-sensitive BI queries, Alvin manages BigQuery reservations at the second level. The agent scales capacity to meet workload demand, ensuring queries stay within established SLOs while minimizing spend on underutilized slots.
  ‍

The Results

• 55% reduction in total BigQuery compute spend for Looker workloads.
• 2x increase in the cache hit ratio, delivering near-instant, zero-cost results.
• Maintained mission-critical performance, ensuring zero impact on user experience or dashboard responsiveness. 
• Improved predictability of daily spend by smoothing out scanning-related spikes.
• Zero maintenance overhead, as Alvin operates autonomously without manual tuning.

"Integrating Alvin was a turning point for our data platform. We achieved a 55% reduction in our Looker spend, but the real value is the automation—ensuring we always optimize both performance and cost for our most time-sensitive analytics. It just works in the background."

‍Doug Parker, Senior Director of Engineering at Flashpoint‍

‍

Billing Model Optimization

Alvin continuously evaluates the computational profile of every query executed by Flashpoint to identify the optimal execution path. For scan-heavy queries, Alvin routes them to a capacity reservation to bypass expensive per-byte charges. For compute-intensive queries with small data footprints, it retains on-demand billing to avoid consuming expensive slot-seconds unnecessarily.

Query Rewriting & Caching

Alvin maximizes cache utilization by rewriting non-deterministic SQL, which doubled the cache hit ratio. By inlining functions like CURRENT_TIMESTAMP(), Alvin ensures that identical dashboard refreshes are served directly from BigQuery’s cache rather than re-executing the query. This reduces compute consumption and provides a faster experience for end-users.

Automated Reservation Management

Beyond routing, Alvin manages Flashpoint’s slot capacity in real-time via API calls to ensure performance standards are never compromised. Utilizing the Autotamer feature, Alvin scales BigQuery reservations at the second level to meet the demand of the routed queries, effectively eliminating performance risks associated with slot contention. This functionality ensures that Flashpoint’s time-sensitive BI dashboards remain highly responsive, automatically scaling capacity to meet stringent SLOs even as Looker demand spikes, while simultaneously minimizing spend on idle compute. 

‍

Ready to Optimize Your Data Spend?

Discover how Alvin can transform your cloud data costs. Schedule a personalized demo or run a free savings analysis to quantify your potential outcomes and see immediate value.