Ballista: Making DataFusion Applications Distributed

Ballista is a distributed query execution engine that enhances Apache DataFusion by enabling the parallelized execution of workloads across multiple nodes in a distributed environment.

Existing DataFusion application:

use datafusion::prelude::*;

#[tokio::main]
async fn main() -> datafusion::error::Result<()> {
  let ctx = SessionContext::new();

  // register the table
  ctx.register_csv("example", "tests/data/example.csv", CsvReadOptions::new())
      .await?;

  // create a plan to run a SQL query
  let df = ctx
      .sql("SELECT a, MIN(b) FROM example WHERE a <= b GROUP BY a LIMIT 100")
      .await?;

  // execute and print results
  df.show().await?;
  Ok(())
}

can be distributed with few lines of code changed:

[!IMPORTANT]
There is a gap between DataFusion and Ballista, which may bring incompatibilities. The community is actively working to close the gap

use ballista::prelude::*;
use datafusion::prelude::*;

#[tokio::main]
async fn main() -> datafusion::error::Result<()> {
    // create SessionContext with ballista support
    // standalone context will start all required
    // ballista infrastructure in the background as well
    let ctx = SessionContext::standalone().await?;

    // everything else remains the same

    // register the table
    ctx.register_csv("example", "tests/data/example.csv", CsvReadOptions::new())
        .await?;

    // create a plan to run a SQL query
    let df = ctx
        .sql("SELECT a, MIN(b) FROM example WHERE a <= b GROUP BY a LIMIT 100")
        .await?;

    // execute and print results
    df.show().await?;
    Ok(())
}

For documentation or more examples, please refer to the Ballista User Guide.

Architecture

A Ballista cluster consists of one or more scheduler processes and one or more executor processes. These processes can be run as native binaries and are also available as Docker Images, which can be easily deployed with Docker Compose or Kubernetes.

The following diagram shows the interaction between clients and the scheduler for submitting jobs, and the interaction between the executor(s) and the scheduler for fetching tasks and reporting task status.

See the architecture guide for more details.

Performance

We run some simple benchmarks comparing Ballista with Apache Spark to track progress with performance optimizations. These are benchmarks derived from TPC-H and not official TPC-H benchmarks. These results are from running individual queries at scale factor 100 (100 GB) on a single node with a single executor and 8 concurrent tasks.

Overall Speedup

The overall speedup is 2.9x

Per Query Comparison

Relative Speedup

Absolute Speedup

Getting Started

The easiest way to get started is to run one of the standalone or distributed examples. After that, refer to the Getting Started Guide.

Cargo Features

Ballista uses Cargo features to enable optional functionality. Below are the available features for each crate.

ballista (client)

ballista-core

ballista-scheduler

ballista-executor

Usage Examples

# Build with standalone support (default)
cargo build -p ballista

# Build with Substrait support
cargo build -p ballista-scheduler --features substrait

# Build with Spark compatibility
cargo build -p ballista-executor --features spark-compat

Project Status

Ballista supports a wide range of SQL, including CTEs, Joins, and subqueries and can execute complex queries at scale, but still there is a gap between DataFusion and Ballista which we want to bridge in near future.

Refer to the DataFusion SQL Reference for more information on supported SQL.

Contribution Guide

Please see the Contribution Guide for information about contributing to Ballista.