gcp Google Cloud Blog ·

BigQuery Enhances Price-Performance for Agentic Workloads

blogdataanalyticsgcpengineermediagcp-bigquery
feature patch

BigQuery has introduced significant price-performance improvements designed to handle the increasing demands of agentic workloads. These updates leverage autonomous query processing, including history-based optimizations and an advanced runtime engine, to automatically tune query performance and reduce costs without user intervention. The enhancements benefit all users, from individual developers to large enterprises, and are available now.

  • BigQuery's Self-Learning Engine with History-Based Optimizations (HBO)
  • Advanced Runtime Engine for Enhanced Query Execution
  • Consistent Performance Across Data Formats
  • Fluid Scaling Autoscaler for Variable Workloads
Features (2)
  • BigQuery's Self-Learning Engine with History-Based Optimizations (HBO)

    BigQuery now features a self-learning engine that utilizes history-based optimizations. This system learns from past query executions to automatically apply beneficial optimizations for future runs, enhancing performance without requiring manual user intervention or SQL modifications. It includes built-in safety guardrails that reject optimizations if they do not improve performance.

  • Advanced Runtime Engine for Enhanced Query Execution

    An upgraded execution engine, BigQuery advanced runtime, automatically determines the optimal physical execution path for queries. This includes enhanced vectorization leveraging new processor efficiencies and optimized handling of 'short' queries to reduce distributed processing overhead and improve latency for high-concurrency applications.

Enhancements (2)
  • Consistent Performance Across Data Formats

    BigQuery's performance improvements are now available regardless of the underlying data format, including open lakehouse architectures using Apache Iceberg. This means users benefit from autonomous acceleration whether using BigQuery's native format or open formats like parquet.

  • Fluid Scaling Autoscaler for Variable Workloads

    The BigQuery autoscaler has been enhanced with fluid scaling, enabling it to manage highly variable workloads efficiently. This feature allows for a premier autoscaling model without compromising performance for cost, directly translating autonomous performance gains into cost savings.

Read the original announcement →

https://cloud.google.com/blog/products/data-analytics/bigquery-performance-optimizations/

Related releases