Delivering Scalable Data Lake Analytics Solutions on Ceph with High Capacity Flash
summary
Discusses using Ceph-based data lake storage combined with high-capacity NVMe QLC NAND flash to meet petabyte-scale analytics needs, describing an all‑flash Ceph architecture and how it enables independent scaling of compute and storage. Presents performance benchmark results for analytics workloads (for example TPC-DS), provides best practices for deployment, and summarizes related work in the Ceph community.
AI-written summary · about
chapters (13)
- 0:00Intro
- 2:35Hadoop/YARN Architecture
- 9:05NVM Express (NVMe)
- 13:00Challenges with Scaling Apache Hadoop
- 15:45Solution - Disaggregated Analytics with Ceph Object Storage
- 22:15QCT Lab Test Environment-System Configuration
- 24:11QCT Lab Test Environment - Setup
- 25:30Test Methodology
- 31:17Test Results - Hive (Hadoop/HDFS vs Hadoop/Ceph)
- 33:30Test Results - Presto (Hodoop/HDFS vs Hadoop/Ceph)
- 34:572019 Plans and Next Steps
- 41:01Summary
- 42:25Call to Action
