Flight schedule calculation 65 percent faster on open-source Java
A global flight information company had hit scaling limits in its schedule calculation system, a mix of DB2 in-database processing and proprietary Java, constrained by both design and licence cost. A short series of proof-of-concept iterations on an open-source high-performance framework, tested against the same production input, cut calculation time by 65 percent.
- -65%
- calculation time against the production system on identical input dataComparative performance test
- Same output
- functional tests confirmed the daily processing output matched production before performance was comparedProof of concept
- Lower TCO
- open-source framework in place of licensed in-database processingProof of concept
- Client
- A global flight information company
- Sector
- Travel: aviation data and schedule processing
- Engagement
- Proof of concept in short spike iterations; a small team of Stratoflow consultants and the client's development lead
- Stack
- Java; an open-source high-performance processing framework replacing DB2 in-database logic and proprietary Java
- Offer today
- Performance Rescue
Scaling limited by design and by licence cost at the same time
The existing schedule calculation system mixed DB2 in-database processing with proprietary Java code. Performance scalability was limited by the system's design and, separately, by commercial licence costs: the obvious way to add capacity also added licence fees. The client needed to know whether a different design could do the same work faster and cheaper before committing to a rebuild.
Prove it on production data before proposing a rewrite
Stratoflow proposed replacing the major calculation logic with an open-source, high-performance framework, and planned a short series of spike iterations to design and implement the selected logic. Comparative performance tests would use exactly the same input data as the production system. A small team was formed from Stratoflow consultants and a development lead from the client.
Functional tests came first: the proof of concept had to produce the same daily processing output as production, so the performance comparison would be like for like.
Three conclusions from the comparative tests
- -65%
- calculation time: the proof-of-concept code was significantly faster than productionComparative test
- Lower
- total cost of ownership with the open-source technology than with the existing licensed mixComparative test
- Simpler
- to develop, maintain and deploy than a system built on an obsolete technology mixEngineering assessment
The offers this engagement proves
Book a scoping call
Thirty minutes with an architect, not a salesperson. You leave with a written view of scope, price model and whether we are the right team for it.