Fresher insights, faster decisions: talabat’s near-real-time analytics across AWS and Google Cloud

This article describes how talabat, a leading delivery app in the MENA region, built a hybrid multi-cloud lakehouse architecture to enable near-real-time analytics across AWS and Google Cloud without duplicating data.

• Streaming events from Apache Kafka on Amazon MSK are processed by Amazon EMR Serverless Spark jobs and written once to Apache Iceberg tables in Amazon S3 Tables on AWS
• BigQuery on Google Cloud reads the AWS-resident Iceberg data through a Lakehouse federated REST catalog, eliminating cross-cloud data movement on the write path
• Cross-cloud access is secured by IAM trust relationships using OIDC federation, with the Lakehouse service account assuming a least-privilege read-only role
• Metadata synchronizes every ~5 minutes from AWS Glue Data Catalog to the federated catalog, making data queryable within minutes of production
• Single physical Iceberg dataset is readable by BigQuery, Amazon Athena, and Spark without duplication, reducing data-movement costs by approximately 40 percent
• Open Apache Iceberg format prevents vendor lock-in and enables multiple compute engines to query the same tables

The architecture keeps the write path short and local on AWS while moving cross-cloud concerns to the read path, expressed through open standards and identity rather than data replication.