将旧版 SAS 代码迁移到 Databricks Lakehouse：我们在此过程中学到了什么.pdf

1、Migrating Legacy SAS Code to Databricks Lakehouse:What We Learned Along the WayMatt Adams(PacificSource)Dmitriy Alergant(T1A)June 10,2025All opinions are our ownA not-for-profit health insurance companyPacificSourceMatt AdamsSenior Data Platforms DeveloperDatabricks Platform LeadDmitriy AlergantPrin

2、cipal Architect at T1APacificSourceWe were founded in 1933.Back in 1945,we insured 4,500 members with a monthly premium of$2!More than half our existence has been without any computers or electronic data processing!https:/Helping people get the healthcare they needYou may call us a health insurance

3、carrier,but theres more to it.Were a not-for-profit family of companies,employing more than 1,950 people who serve over 500,000 members and 3,900 businesses across Idaho,Montana,Oregon,and Washington.3A not-for-profit health insurance companyFrom Legacy ETL to DatabricksAccelerate core reporting dat

4、a models from batch to real-timeprocessing using DLT pipelines4Our Transition PathDatabricks platform deployment,exploration,and migration planning 2022 2023-2024 Migration of Core and Auxiliary data models from legacy ETL tools such as SQL Server,SSIS,Informatica to Databricks Lakehouse(EDW)Impleme

5、nted full Azure DevOps CI/CD build pipelines,using Databricks Asset Bundles(DAB)Reports availability SLAs went from 0-4%to 95%2025&BeyondMigration of SAS Data Mart Legacy DWH Data SourcesMembership and Groups DimensionsPremiumsRaw Claims Facts Pre-ProcessingDental ClaimsVision ClaimsRX Drugs ClaimsP

6、hysician ClaimsOutpatient ClaimsInpatient ClaimsAll Claims aggregation and Member Cumulative ReportingActuarial Reserves CalculationSAS Analytical Data MartLegacy State70 scripts 25,000 lines of code with very complex business logicTakes 4 days to refresh each month,with manual interventionsSAS used