不断演进的列车线路架构以实现规模化、可靠性和生产力.pdf

1、Lessons from scaling Trainlines architectureMilena Nikolic,CTO13 April 2024Private|Confidential213 April 2024Private|Confidential35B5B net ticket sales350350 searches per second over 3.8M monthly unique routes500500 Tech and Product people13 April 2024Private|Confidential4Whats difficult about what

2、we doAggregation of supplyTransactions over a finite inventoryTicket fulfillment speed13 April 2024Private|Confidential5270 API 270 API integrationsintegrations100 x more 100 x more train train stations stations than than airportsairports1300 1300 transactions transactions per min at per min at peak

3、peakAccess Access pattern pattern restrictionsrestrictionsFares,Fares,classes,classes,optionalsoptionalsSignificant Significant percentage percentage of tickets of tickets sold on the sold on the day or travelday or travelBarcode Barcode expected expected instantlyinstantly3 lessons on scaling Scali

4、ng teams and scaling tech to scale productivity and impactStory 1:Scaling team productivity13 April 2024Private|Confidential7Teams shaped around the ownership of the parts of technical stack,never owning any business goals e2e.Platform owns all the tech surface and is on the hook for sustain work an

5、d tech strategy.Verticals aligned to product and business goals,contributing to the tech stack owned by the platform.A set of core shared services owned by the Platform(50%of the overall tech surface),the remainder aligned to the most relevant vertical.Clusters(pre 2022)Team structure and stack owne

6、rship evolutionPlatform and Verticals(2022-2024)Core Platform and Diagonals(current)APQAPQAPQEach structure gives different answers to two key questions:1.Who owns what part of technical surface?2.Which product/business/tech goals is team on the hook for?Teams shaped around the ownership of the part