What is ML infrastructure article card image

How to Build Great Machine Learning Infrastructure

Ray glossary image

Ray is an open source AI Compute Engine for scaling AI and Python applications. Build and run distributed applications better with Ray. 

Map Small

What is Ray? 

Ray Data glossary image

Learn how to get started with Ray Data, an open source distributed training library. Built on Ray Core, Ray Data enables data processing ML workloads at scale.


Ray Data Streaming for Large-Scale ML Training and Inference

Fast, Flexible, and Scalable Data Loading for ML Training with Ray Data

What is the Ray Data Library?

Ray Serve glossary image

Ray Serve is a scalable model serving library for building online inference APIs.

Deploying Many Models Efficiently with Ray Serve

Enabling Cost-Efficient LLM Serving with Ray Serve

What is Ray Serve? 

Ray Train glossary image

Ray Train: A Production-Ready Library for Distributed Deep Learning

What is Ray Train? 

Ray vs Spark glossary image

Explore why Ray is better at supporting the future of AI/ML workloads thanks to heterogeneous compute and its best-in-class unstructured data processing.

Apache Spark was built for a world that was CPU centric and relied on structured and semi-structued data. Ray is better at supporting the future of AI/ML workloads because of its ability to enable heterogeneous compute (compute across GPUs and CPUs) as well as best-in-class unstructured data processing.

With Ray, Amazon could compact 12X larger datasets than Apache Spark, improve cost efficiency by 91%, and process 13X more data per hour.

Amazon White

The first thing to understand when comparing Spark and Ray for specific data processing use cases is that you’re not actually comparing Spark and Ray. 
<br/>
__Spark__—though it has select support for traditional machine learning through libraries like MLlib—is primarily a big data processing tool.
<br/>
__Ray__ as a whole is an open source framework for distributed computing. Built on top of Ray Core are several high-level libraries that focus on specific AI workloads—including data processing with __Ray Data__.
<br/>
In order to directly compare Spark and Ray for data processing, we will be comparing Apache Spark to Ray Data in the following section. 

Ray is the most powerful open source distributed computing framework. Anyscale, built by the creators of Ray, takes Ray to the next level. RayTurbo, Anyscale’s optimized Ray, builds on Ray and expands it to make it even more powerful, performant, and reliable. 

Anyscale light rebrand

What is the Difference Between Ray and Apache Spark?

Any accelerator, any cloud, any framework. Anyscale is made for how you work today<br class="hidden lg:block"/>and how you want to scale tomorrow.

Glossary

Call for Papers for the vLLM Featured Track at Ray Summit is now Open!

Anyscale is the leading AI application platform. With Anyscale, developers can build, run and scale AI applications instantly.