Clipper DB Accelerator for Spark
Proven Performance and Cost Saving at Scale
Spark Applications
SQL
DataFrames
ETL
ML
BI
Notebooks
Spark Query Plans
Catalyst Optimizer
Native Worker 1
Fully Native Execution
ClipperDB Accelerator™
Fully Native Execution™
Precise Parallel Prefetching™
Exploit Cloud Bandwidth to Fully Utilize vCPUs
Dynamic Data Caching
Cache prefetched data in cluster compute nodes for faster query performance
In-Memory Row Group Processing™
Optimize throughput of native batch pipelines
Parallel Dataflow Native Pipelines™
Increase throughput and resource utilization
Streaming No-Copy Exchange
High-speed, zero-copy data exchange between workers
Cloud Store Checkpoint Query Fault Tolerance™
Efficient recovery for long-running jobs
Google Cloud Storage
Amazon S3
Apache Iceberg Tables
Parquet Data Files
Other Cloud Data Lakes
Native Worker N
Fully Native Execution
Customer Outcomes
Speedup jobs and queries up to 10x+
Reduce $/query 4x+
Bootstrap action, transparent acceleration
Smaller Clusters, No JVM headaches
Built for mission-critical workloads