Next-Gen Shuffle Service

A high-performance Rust implementation of Apache Uniffle, delivering unparalleled performance for large-scale data processing workloads in your infrastructure.

production ready
50% faster than Spark ESS
infinitely scalable
Get Started View on GitHub

Set it and forget it

Built on Apache Uniffle

Leveraging the power of Riffle's Rust implementation to provide superior shuffle service performance.

Superior Performance

Up to 50% faster than vanilla Spark ESS with optimized URPC implementation.

Flexible Storage

Support for Memory, LocalFile, and HDFS with intelligent hybrid storage options.

Performance that speaks for itself

50%
Faster than vanilla Uniffle
3.2min
TeraSort 273G Data
2000+
Concurrent Spark Jobs

Why Riffle?

Traditional Apache Uniffle

  • Standard performance
  • Java-based implementation
  • Basic storage options
  • Periodic GC pauses
  • Fluctuating memory usage
  • Traditional I/O operations

Riffle Advantages

  • 50% faster performance and lower latency
  • Rust-powered efficiency
  • Advanced hybrid storage
  • No JVM GC pauses
  • Predictable memory consumption
  • Direct I/O for improved performance and lower CPU usage