重新设计 Kaizen 的未来云数据湖.pdf

1、Redesigning Kaizens Cloud Data Lake for the FutureNikolaos Michail,Triantafyllos Tsakmakis June 10,2025About Us3AgendaIntroductionData in Kaizen GamingData Platform ChallengesArchitectureCloud Adoption FrameworkData Landing ZonesOld vs New architecture Challenges RevisitedMigration Phases break down

2、Infrastructure&CodeData migrationConclusions GameTech companyBetano Brand in 18 markets13M+active users1.5M+Gaming events per year5Data in Kaizen 2 PetaBytes in Data LakeLakehouse Azure DBxDaily:400M TxMinute:700K Tx(peak)From 27M to 121B in 10 years(+4,500%)6Data Teams Kaizen Data DNA150+Headcount

3、Data&BI&AI 85 Departments served 1k+Internal End UsersData Mesh Paradigm(Domains)7Data Platform ChallengesTechnical Debt accumulationGood enough decisionsdo not fit our scale Quotas(Limits)Difficult SDLCNo cost transparencyGovernance Value vs Optimization Shared ResourcesChange managementError-prone

4、 environment provisioningInfrastructure8GoalsScalabilitySecurityCost TransparencyCost Efficiency Opportunity9Implicit GoalsDomain boundariesData democratizationCloud MigrationCloud ArchitectureCloud Adoption Framework(CAF)Reference Cloud ArchitectureCAF Implementation on AzureData(Management)Landing

5、 Zones Landing Zones Terraform Module11Azure ImplementationCloud Resource HierarchyTenant:Identity ManagementSubscription:Quotas,Billing,IsolationResource Group:Logically RelatedResources:Workspace and Storage12Azure TerminologyOld ArchitectureCommon QuotasSingle Storage TypeManually Created Resourc

6、esFinOpsDifficult to Scale13ChallengesEnvironments/Landing Zones SeparationCapacity Planning:Workspaces,NetworkResource Synergy Resource Properties per use caseFuture proofingDesign14ConsiderationsNew ArchitectureLanding ZonesNew ArchitectureEnvironmentsArchitecture Goals RevisitedCost TransparencyC