Slash Apache Spark Cost per Query 4x
Data Redefined
ClipperDB Accelerator transparently replaces Spark JVM executors with a fully native query execution engine while preserving Spark compatibility
4x
11x
>2x
lower cost per query
up to 11x faster analytics
performance on half-size clusters
Same Analytics. Twice the Performance on Half the Infrastructure.
Based on patented inventions
Precise Parallel Prefetching
Dynamic Cloud Caching
Parallel Dataflow Pipelines
In-Memory Row Group Processing
Streaming Exchange
Query Fault Tolerance
Execute Spark query plans on fully native workers, eliminating JVM overhead and maximizing compute efficiency. Precise Parallel Prefetching™ exploits cloud bandwidth to keep vCPUs fully utilized, while Dynamic Cloud Data Caching stores prefetched data in cluster memory for consistently fast access. Parallel Dataflow Native Pipelines™ with Parallel Dataflow Native Pipelines™ optimize vCPU throughput, and Streaming Exchange™ moves data between workers without materialization or memory copies. Cloud Store Checkpoint Query Fault Tolerance™ enables efficient recovery from failures, allowing long-running jobs to resume with minimal recomputation. Together, these patented innovations deliver faster analytics, greater resilience, and lower infrastructure costs at scale.
Uses Proven Open-Source Technology
VELOX
Apache Spark
NATIVE EXECUTION LIBRARY
Apache Spark™
Patents
Disaggregated query processing utilizing precise, parallel, asynchronous shared storage repository access
U.S. Patent No. 11,327,966 B1:
Disaggregated query processing on data lakes based on pipelined, massively parallel, distributed native query execution on compute clusters utilizing precise, parallel, asynchronous shared storage repository access
Amazon EMR Accelerator
