Skip to content

Latest commit

 

History

History
26 lines (20 loc) · 1.83 KB

File metadata and controls

26 lines (20 loc) · 1.83 KB

Executive Summary

The DBA CPU Benchmark1 report is a synthetic benchmark designed to measure the single-threaded CPU performance of the database server. It generates a large dataset (500,000 rows) by creating a Cartesian product of standard Oracle EBS tables (FND_LANGUAGES, FND_CURRENCIES, FND_APPLICATION). The execution time is dominated by the PL/SQL engine's ability to process these rows and generate the output, making it a good proxy for PL/SQL performance.

Business Challenge

  • Hardware Validation: "We migrated to new servers. Are the CPUs actually faster?"
  • Cloud Migration: "How does the performance of an AWS r5.xlarge compare to our on-premise Exadata?"
  • Baseline Creation: Establishing a performance baseline before applying OS patches or upgrades.

Solution

This report runs a query that forces the database to do significant work (generating rows) and measures the elapsed time.

Key Features:

  • Standardized Workload: Uses standard FND tables present in every EBS instance, ensuring comparability.
  • CPU Bound: Designed to be CPU-intensive (assuming the data is in the buffer cache), minimizing I/O noise.
  • Comparative Data: The description includes reference timings for various processors (e.g., AMD Ryzen, Intel Xeon).

Architecture

The report queries FND_LANGUAGES, FND_CURRENCIES, and FND_APPLICATION.

Key Tables:

  • FND_LANGUAGES, FND_CURRENCIES, FND_APPLICATION: Used as data sources for the Cartesian product.

Impact

  • Vendor Accountability: Verifies that the cloud provider or hardware vendor is delivering the promised performance.
  • Sizing Decisions: Helps determine the correct CPU SKU for a new deployment.
  • Performance Troubleshooting: Rules out (or confirms) "slow CPU" as a cause of general system sluggishness.