Run Apache Spark up to 10x faster with DataPelago on Amazon EKS

This article explains how DataPelago Accelerator for Spark (DPA-S) accelerates Apache Spark workloads on Amazon EKS, delivering significant performance and cost improvements.

• DPA-S achieves up to 10x faster execution and 80% cost reduction without code changes or data migration
• DataPelago Nucleus engine intercepts Spark physical plans and routes operations to CPUs or GPUs based on runtime data analysis
• Outperforms NVIDIA cuDF with 3.7-38.6x higher throughput on projections, filtering, aggregations, and string processing
• Deploys as a Spark plugin with seamless integration into existing Amazon EKS clusters and security configurations
• Production customers report 2-10x performance gains and 50-80% cost reduction across social media, analytics, ecommerce, and financial services
• Available in AWS Marketplace with proof-of-concept deployments typically completed in 3-4 weeks

DataPelago enables enterprises to leverage heterogeneous compute infrastructure for Spark workloads without rewriting applications or managing complex GPU optimization.