From 30rpm to 1980rps

How did we change the mindset of a team that used to work an absurd legacy code. Recently, they rewrote a software into a new one able to process 1980 requests per second.

Transcript

1. from 30rpm to 1980rps The “humble but hungry” team power
2. Original Software • COBOL based business logic and storage • External access to information through a socket server written in C • Totally un-modular • Hundreds of thousands of IF-ELSE's Lines of Code • 0% Code Coverage
3. Original Software performance • Not scalable • 30 RPM • 20MB ~ 100% RAM usage • 15% CPU usage • a bug makes log file grown about a 1GB per minute
4. Why should it be rewritten? • Original C socket server "works luckily" • Several performance issues • Several customization points avoid new customizations to be developed in affordable time
5. Why should it be rewritten? • There was one (SEVERE) bug that appears randomly under production once in a while • There was a new project soon that will bring new customizations
6. New software • COBOL based business logic and storage • Java multi-protocol Server • Modular customizations • Orthogonal customizations
7. New software performance • Vertically/Horizontally scalable • 1980 RPS ( due to a limitation on COBOL piped communication ) • 85% code coverage
8. how did we achieve this results in three months?
9. how did we achieve this results in 3 months?
10. we’ve built a humble, but hungry, team!
11. "Win a discussion" vs "Discuss to conclude something"
12. Benefits Driven
13. Be open to try something new
14. On-demand code review
15. development boundaries
16. Max Lines of Code: 9/200
17. Focus on "SO" from SOLID
18. Minimum Viable Design
19. Continuous Research and Development
20. Zero Rework Policy
21. Feedback tests on Business Behaviors
22. Questions?
23. thanks! E-mail: miere.teixeira@texoit.com E-mail: miere.teixeira@gmail.com Skype: miere.teixeira Twitter: miere GitHub: miere