GOAI: GPU Open Analytics Initiative

Brad Nemire

Continuum Analytics, H2O.ai, and MapD recently announced the formation of the GPU Open Analytics Initiative (GOAI) to create common data frameworks enabling…

NVIDIA

•

Brad Nemire

•2 min read•beginner•

--

•View Original

H2O.aiJavaPandasPythonScalaSQL

Overview

The GPU Open Analytics Initiative (GOAI) aims to create common data frameworks that enhance data science on GPUs, enabling seamless data interchange among applications. This initiative, supported by key players like Continuum Analytics, H2O.ai, and MapD, focuses on improving the efficiency of data analytics by leveraging the high throughput capabilities of GPUs.

What You'll Learn

1

How to utilize the GPU Data Frame API in Python for machine learning tasks

2

Why standard analytics data formats are crucial for GPU-accelerated data science

3

How to achieve significant performance improvements in data analytics using GPUs

Prerequisites & Requirements

Familiarity with traditional big data tools like Hadoop and Spark
Understanding of data science languages such as Python, SQL, and R

Key Questions Answered

What is the purpose of the GPU Open Analytics Initiative (GOAI)?

The GPU Open Analytics Initiative (GOAI) aims to create common data frameworks that facilitate seamless data interchange among applications, thereby enhancing the efficiency of data science on GPUs. This initiative is expected to contribute significantly to the growth of GPU computing in accelerated analytics.

How does GOAI improve data analytics performance?

GOAI improves data analytics performance by providing standard analytics data formats for GPUs, which reduces excessive data movement and conversion costs. This allows developers to fully utilize the high throughput capabilities of GPUs, resulting in faster data processing.

What performance improvements can be achieved using GOAI?

Using GOAI's GPU Data Frame API, significant performance improvements can be achieved, such as a reported 35x speedup when running a Generalized Linear Model on US Census data stored in a MapD database, compared to a dual-Xeon CPU-only system.

Key Statistics & Figures

Performance speedup

35x

Achieved by running a Generalized Linear Model on US Census data using 8 Tesla P100 GPUs compared to a dual-Xeon CPU-only system.

Technologies & Tools

API

GPU Data Frame API

Used to perform machine learning algorithms on data stored in a MapD GPU-accelerated database.

Database

Mapd

A GPU-accelerated database used for storing and processing data in conjunction with GOAI.

Machine Learning Framework

H2o.ai

Utilized for running machine learning algorithms within the GOAI framework.

Key Actionable Insights

1
Leverage the GPU Data Frame API to enhance your machine learning workflows.
By using the GPU Data Frame API, data scientists can directly share data on GPUs, which minimizes data movement and maximizes processing speed, making it ideal for large datasets.

2
Adopt standard analytics data formats to streamline your data science processes.
Implementing standard formats can significantly reduce the overhead associated with data conversion and movement, allowing for more efficient analytics workflows on GPUs.

3
Explore the potential of GPU computing in your data analytics projects.
Utilizing GPU computing can lead to drastic improvements in performance, as evidenced by the 35x speedup achieved in specific machine learning tasks, making it a valuable consideration for data-intensive applications.

Common Pitfalls

1

Neglecting the importance of data format standardization can lead to inefficiencies.

Without standardized data formats, workflows may suffer from excessive data movement and conversion costs, which can hinder performance and slow down analytics processes.