Project Tungstun

Overview

Based on our observation, majority of Spark workloads are not bottlenecked by I/O or network, but rather CPU and memory. This project focuses on 3 areas to improve the efficiency of memory and CPU for Spark applications, to push performance closer to the limits of the underlying hardware.

Memory Management and Binary Processing

• Avoiding non-transient Java objects (store them in binary format), which reduces GC overhead.
• Minimizing memory usage through denser in-memory data format, which means we spill less.
• Better memory accounting (size of bytes) rather than relying on heuristics
• For operators that understand data types (in the case of DataFrames and SQL), work directly against binary format in memory, i.e. have no serialization/deserialization

Cache-aware Computation

• Faster sorting and hashing for aggregations, joins, and shuffle

Code Generation

• Faster expression evaluation and DataFrame/SQL operators
• Faster serializer

Several parts of project Tungsten leverage the DataFrame model, which gives us more semantics about the application. We will also retrofit the improvements onto Spark’s RDD API whenever possible.

Deep Dive into Project Tungsten Bringing Spark Closer to Bare Metal

Josh Rosen (Databricks)

From DataFrames to Tungsten A Peek into Spark’s Future

Reynold Xin (Databricks)