AI innovation passes a major airline's ultimate ETL test

Met and beat Prior environment performance benchmarks under real workloads
Platform-wide Enhancements benefit all future Compute Engine customers
Expanding Additional workloads beyond fare calculation

はじめに

概要

A strategic collaboration that delivered—and strengthened Teradata's platform for every future customer

A major airline wanted to modernize large-scale fare calculation exchange, transform, and load (ETL) workloads using Teradata's platform, bringing on-demand compute flexibility without sacrificing the strict performance windows their operations require. What followed was an intensive cross-functional engineering collaboration that not only delivered results for the airline but also advanced Teradata's core Compute Engine architecture for the benefit of all future customers.

Airline needed on-demand compute flexibility for time-critical ETL without sacrificing performance
Teradata stress-tested under real-world, large-scale fare calculation workloads
Engineering and AI Services collaborated daily with full leadership alignment throughout
Platform enhancements made during the engagement benefit all future Compute Engine customers

チャレンジ

Making separated compute and storage perform like co-located infrastructure under real deadline pressure

The airline’s existing environment kept data physically close to processing systems, enabling fast, predictable ETL performance within tightly defined operational windows. Teradata introduced a different architectural model: on-demand compute with separately managed storage. This flexibility and cost efficiency came with a performance challenge—the distance between compute and data introduced latency that needed to be engineered away. Internal debate within the airline added further complexity, as a subset of stakeholders required intermediate ETL steps to be preserved for process consistency.

Separated compute and storage introduced latency not present in the airline's existing co-located architecture
ETL workloads run on fixed operational schedules—no room for performance shortfalls
Internal stakeholder requirements mandated restoration of intermediate steps after initial optimization
Engineering had to deliver both architectural integrity and customer process preferences simultaneously

ソリューション

Line-by-line SQL analysis, ETL refactoring, and deep platform tuning: A true 360-degree collaborative effort

Teradata's engineering and AI Services teams embedded with the airline in a daily operating rhythm—with internal alignment spanning engineering, product, support, account teams, and leadership. Engineering analyzed SQL queries line by line to identify latency points. AI Services refactored ETL logic to balance architectural integrity with the airline’s process requirements. When incremental tuning proved insufficient, teams went deeper into platform behavior, making fixes at both the workflow and platform level. A strict 360-degree lens ensured every solution worked across all functions, not just in isolation.

SQL analyzed line by line to identify and eliminate latency points in the separated compute model
ETL logic refactored collaboratively to balance performance with stakeholder process requirements
Deep platform behavior fixes made where workflow-level tuning was insufficient
360-degree review ensured changes were sound across engineering, support, and account dimensions

結果

Performance that removed doubt—and a platform strengthened for every future customer

Teradata consistently met—and in many cases exceeded—the airline's prior environment performance benchmarks, validating the separated compute-and-storage architecture under real operational conditions. Internal debate within the airline was settled by the results. The airline is now expanding to additional workloads. And every enhancement made during this engagement strengthens the Teradata platform for future customers.

Met and beat: Teradata matched and exceeded the airline’s prior environment performance under real ETL workloads
Expanding use: Airline moving beyond fare calculation to explore additional workloads on the platform
Platform-wide gains: Enhancements made for the airline improve execution and tuning for all future Compute Engine customers