From inception to insight: Accelerating AI productivity ...

From inception to insight: Accelerating AI productivity with GPUs

John Zedlewski, Director, RAPIDS Machine Learning @ NVIDIA Ramesh Radhakrishnan, Technologist @ Server OCTO, Dell EMC

Data sizes continue to grow

Prototyping and production diverge

Large-scale cluster

Spark, Hadoop High throughput

Full data

Workstation

Python Fast iteration Small data subset

Challenges!

"Tools gap" - rewriting Python or R code to Spark/Hadoop jobs to scale to cluster

High latency on cluster leads to slower iteration

Small data subsets on workstation make it hard to build realistic models

RAPIDS on GPU

RAPIDS + Dask

Consistent tools: Workstation or Cluster

High throughput / low latency

Full data or large subsets

2

Data Processing Evolution

Faster data access, less data movement

Hadoop Processing, Reading from disk

HDFS Read

Query

HDFS HDFS Write Read

Spark In-Memory Processing

ETL

HDFS HDFS Write Read

HDFS Read

Query

ETL

ML Train

Traditional GPU Processing

HDFS Read

GPU Read

QueryWCrPiUte

GPU Read

ETL

CPU Write

GPU ML Read Train

5-10x Improvement More code

Language rigid Substantially on GPU

ML Train

25-100x Improvement Less code

Language flexible Primarily In-Memory

3

DDaattaa MMoovveemmeenntt aanndd TTrraannssffoorrmmaattiioonn The bane of productivity and performance

APP B

Read Data

CPU

APP B Copy & Convert APP A

Copy & Convert Copy & Convert

APP B

GPU Data

GPU

GPU Data

APP A

APP A

Load Data

4

DDaattaa MMoovveemmeenntt aanndd TTrraannssffoorrmmaattiioonn What if we could keep data on the GPU?

APP B

Read Data

CPU

APP B Copy & Convert APP A

Copy & Convert Copy & Convert

APP B

GPU Data

GPU

GPU Data

APP A

APP A

Load Data

5

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download