Engineering Insights from Pinterest: Customer-Facing Analytics that Pays скачать в хорошем качестве

Engineering Insights from Pinterest: Customer-Facing Analytics that Pays

Pinterest Senior Staff Engineer and Presto TSC member Zhenxiao Luo walks through how StarRocks became a cornerstone of Pinterest’s analytics stack. You'll get an in-depth look at how they approached customer-facing analytics, overcame limitations with legacy tools like Apache Druid and HBase, and optimized their Partner Insights platform—cutting P90 latency by 50% while using just 32% of the previous instances. Dive into this video for Zhenxiao’s insights on Pinterest’s journey! ----------------------------------------------------------------------------------------------------------------------- Timestamps 00:00 Intro 02:53 Agenda 04:20 What Does Pinterest Do 05:37 Analytics at Pinterest: Internal and External Customer-Facing Analytics Use Cases 05:53 Partner Insights for Business Partners 06:30 Pinalytics - Pinterest Internal Analytics 07:23 Experiment Metric Platform 08:16 Guardian - Anti-Spam Platform 09:16 Ads Reporting 10:19 What Are the Challenges With Pinterest’s Previous Solutions 13:34 What Is StarRocks 19:29 What Made Pinterest Switch to StarRocks 22:50 Integrating StarRocks Into Pinterest - How Pinterest Is Making StarRocks Work for Them 26:06 Archmage - The Proxy Layer 29:25 StarRocks for Real-Time Insights: Use Cases 29:36 StarRocks for Partner Insights - Performance Boost and Cost Savings 30:53 StarRocks for Pinalytics - How StarRocks Solves Challenges HBase and Druid Couldn't—And Cut Costs by Millions 33:46 Experiment Metrics on StarRocks - From Presto to StarRocks 34:59 Guardian on StarRocks - Performance Boost and Cost Savings 36:50 Ads Reporting on StarRocks - Simplified Data Pipelines and Cost Savings 39:25 StarRocks’ Share-Nothing Mode & Shared-Data Mode and Why Pinterest Chose Shared-Data Mode 41:45 EBS-Based Backup Restore 42:59 Plans and Next Steps 44:44 Q&A 45:00 Is StarRocks running in shared mode? 45:43 How do you ingest data? 46:26 What is the ad hoc concurrency? 46:57 Did I hear correctly that you moved from shared-data to shared-nothing architecture? If so, could you tell us more about why you chose to do that? 48:06 How do you manage CDC? 48:52 What do you use for a reporting front end for the StarRocks data? 50:00 Is this deployed in Kubernetes or EC2 servers? 50:48 Do you have a layer that translates user queries to the 20k+ underlying tables, if you allow analysts to query the tables directly? 51:59 Did you benchmark external Iceberg performance compared to data in real tables? 53:36 Further Reading: More Customer-Facing Analytics Success Stories ----------------------------------------------------------------------------------------------------------------------- Learn more at https://celerdata.com/ Connect with us: LinkedIn:   / celerdata   Twitter:   / celerdata   CelerData Website: https://celerdata.com/ StarRocks GitHub: https://github.com/StarRocks/StarRocks StarRocks Website: https://www.starrocks.io/ Slack: https://starrocks.io/redirecting-to-s... #DataAnalytics #DataEngineering #RealTimeAnalytics #RealTimeData #OLAP #DataAnalyst #DataEngineer #DataInfrastructure #UserFacingAnalytics #Database #AnalyticalDatabase #Denormalization #DataScience #Presto #ApacheDruid #HBase #CustomerFacing