Tungsten execution engine

E705280

Tungsten execution engine is a low-level, memory- and CPU-optimized execution backend in Apache Spark designed to significantly improve performance of data processing workloads.

Try in SPARQL Jump to: Statements Referenced by

Statements (47)

Predicate Object
instanceOf Spark execution backend
query execution engine
basedOn project Tungsten NERFINISHED
benefits better hardware utilization
higher throughput
lower latency
componentOf Spark SQL Catalyst optimizer pipeline NERFINISHED
designedBy Apache Spark community NERFINISHED
Databricks engineers
developedFor Apache Spark Core NERFINISHED
Apache Spark SQL NERFINISHED
goal improve performance of data processing workloads
implements whole-stage Java code generation for query plans
improves CPU utilization
memory utilization
query execution speed
introducedIn Apache Spark 1.4 NERFINISHED
language Java NERFINISHED
Scala NERFINISHED
manages off-heap memory pages
nameOrigin named after metal tungsten
nameReason emphasizes performance and efficiency
optimizationFocus CPU efficiency
cache locality
memory efficiency
partOf Apache Spark NERFINISHED
reduces Java object allocation
garbage collection overhead
interpretive query execution overhead
relatedTo Spark SQL Catalyst optimizer NERFINISHED
Spark physical execution layer
replaces row-based interpreted execution in Spark SQL
runsOn Java Virtual Machine NERFINISHED
storesDataAs compact binary rows
supports DataFrame API
Dataset API NERFINISHED
Spark SQL queries
targetEnvironment distributed data processing
targetWorkloads ETL pipelines
batch processing
interactive SQL queries
uses binary row format
cache-friendly data structures
expression code generation
off-heap memory management
runtime code generation
whole-stage code generation

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Apache Spark provides Tungsten execution engine