
Gautam Kowshik
Senior Computer Scientist with Platform Query Service Team at Adobe.
Actions
Been interested and working on Large Scale Distributed Systems, Analytics & Open Source Software for over 10 years. Part of the Platform Query Service at Adobe. Early Apache Iceberg Contributor.
Iceberg At Adobe
We were on our second iteration of Adobe Experience Platform (AEP)'s Data Lake when Iceberg first came up on our radar. At the time we were struggling with our internal meta store that was suffering from scalability and consistency issues that we had to solve. We quickly saw the synergy with Iceberg and appreciated its design, which was both lightweight and extensible. Iceberg addressed critical issues for us around data reliability; such as filtering partial results due to failures and data consistency with parallel reads and write. It also provided support for partition pruning with partition ranges and file skipping with column level stats that our queries could benefit from immediately. Iceberg enabled new possibilities for us around Time-Travel, Point In Time Recovery, and Incremental Reads. Finally, Iceberg was open sourced and had a vibrant community evolving it.
The goal of this presentation is to share our journey at Adobe with building a Data Lake on-top of Iceberg. Customers use AEP to centralize and standardize their data across the enterprise resulting in a 360 degree view of their data of interest. That view can then be used with intelligent services to drive experiences across multiple devices, run targeted campaigns, classify profiles and other entities into segments, leverage advanced analytics, and more. As a consequence of this we are processing millions of batches per day (TB of Data), writing to thousands of datasets, and scanning Peta-bytes of data on the Data Lake. The journey wasn't all rainbows and unicorns, rather we did hit some bumps along the way. We will go into details on how we solved problems like high-frequency small files, exactly once writes, support for Adobe's Experience Data Model (XDM) schemas, fine tuning of query optimizations, and tombstoning of data for General Data Protection Regulation (GDPR) prior to the v2 format. Having successfully migrated our customers to Iceberg backed DataSets we can end on a good note and share how Iceberg is performing in Production and what's next with Iceberg at Adobe.
Subsurface LIVE Winter 2021 Sessionize Event
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top