Machine Learning in the Enterprise 2019

MACHINE LEARNING IN THE ENTERPRISE
Timothy Spann | Senior Solutions Engineer
@PaasDev

2 © Cloudera, Inc. All rights reserved.
DISCLAIMER
DA

Introduction
Tim Spann has been running meetups in Princeton on Big Data technologies since 2015.
Tim has spoken at several international conferences on Apache NiFi.
https://community.hortonworks.com/users/9304/tspann.html
https://dzone.com/users/297029/bunkertor.html
https://www.meetup.com/futureofdata-princeton/
https://dzone.com/articles/integrating-keras-tensorflow-yolov3-into-apache-ni

Hadoop {Submarine} Project: Running deep learning workloads on YARN ,
Tim Spann (Cloudera)

IOT EDGE PROCESSING WITH MINIFI AND MULTIPLE DEEP LEARNING LIBRARIES

The Industry’s First Enterprise Data Cloud
From the Edge to AI

WHY CLOUDERA?
One stop shop for analytics
Unified open architecture
Hybrid and multi-cloud
INGEST &
STREAMING
DATA
SCIENCE
DATA
WAREHOUSE
OPERATIONAL
DATABASE
DATA
ENGINEERING

CLOUDERA DATA FLOW (CDF)

MACHINE LEARNING PHASES
Where to Connect to Apache NiFi

HANDS ON
CDSW + NiFi
https://community.hortonworks.com/articles/239961/using-cloudera-data-science-workbench-with-apache.html

CLOUDERA DATA SCIENCE

MACHINE LEARNING IS A GROWTH ENGINE
PROTECT
business
CONNECT
products & services (IoT)
DRIVE
customer insights
●
●
●
●
●
●
●
●
●
It’s enabling entirely new businesses, not just modernizing existing systems.
Machine learning refers to algorithms and methods to extract useful patterns from data.
When we say machine learning, we mean broad, transformational data capabilities.

MOVING FROM EXPLORATION TO PRODUCTION OF ML & AI
WE’RE WITNESSING THE INDUSTRIALIZATION OF AI
FROM THE LAB… TO THE FACTORY

ENTERPRISE-GRADE AI OPERATIONS
WHETHER YOU ARE A FORTUNE 100 OR A STARTUP
SECURITY,
GOVERNANCE,
COMPLIANCE
STRATEGY PEOPLE &
ORGANIZATION
TECHNOLOGY

AI
MACHINE
LEARNING
DATA SCIENCE
ANALYTICS
"BIG DATA"
CLOUD

MACHINE LEARNING AT CLOUDERA
Our philosophy
●
●
●

OUR APPROACH
Modern enterprise platform, tools and expert guidance to help you unlock
business value with ML/AI
Agile platform to build,
train, and deploy many
scalable ML applications
Enterprise data science
tools to accelerate
team productivity
Expert guidance,
services & training to
fast track value & scale

PLATFORM

© Cloudera, Inc. All rights reserved. 24
AND ONE MORE THING….

Amazon
S3
Microsoft
ADLS HDFS KUDU
SECURITY GOVERNANCE
WORKLOAD
MANAGEMENT
INGEST &
REPLICATION
DATA CATALOG
Core
Services
Storage
Services
ANALYTIC
DATABASE
DATA
SCIENCE
EXTENSIBLE
SERVICES
OPERATIONAL
DATABASE
DATA
ENGINEERING
MACHINE LEARNING IS BUILT ON DATA MANAGEMENT
Integrated data, workflows, metadata, security, governance, ...

CLOUDERA
ENTERPRISE
DATA
PLATFORM
The modern platform
for machine learning &
analytics optimized for
the cloud
WORKLOADS 3RD
PARTY
SERVICES
DATA
ENGINEERING
DATA
SCIENCE
DATA
WAREHOUSE
OPERATIONAL
DATABASE
DATA CATALOG
GOVERNANCESECURITY LIFECYCLE
MANAGEMENT
STORAGE
Microsoft
ADLS
COMMON SERVICES
HDFS
Amazon
S3
CONTROL
PLANE
KUDU

CLOUDERA DATA SCIENCE WORKBENCH

ACCELERATING THREE STAGES OF MACHINE LEARNING
Manage models
Deploy models
Monitor performance
DEPLOYDEVELOP
Explore data
Develop models
Share results
TRAIN
Optimize parameters
Track experiments
Compare performance
Enterprise AI platform supporting model development, training, and deployment

A PLATFORM FOR
MACHINE LEARNING
• Open platform
• Complete lifecycle
• Team collaboration
• Enterprise ready
• Runs anywhere
RESEARCH | PRODUCTION
LOCAL | SPARK | IMPALA/HIVE
DEPLOYMENT
COMPUTE
OPEN SOURCE ECOSYSTEMALGORITHMS
SELF-SERVICE
TOOLS
SOLUTIONS | USE CASESAPPS
CLOUD ON-PREMISES
ADLSS3 HDFS KUDU
CATALOG | SECURITY | GOVERNANCE
SHARED
CONTEXT

THE CHALLENGE
Balance these needs
DATA SCIENCE
•Access to granular data
•Flexibility
• Preferred open source tools
•Elastic provisioning
• Compute
• Storage
•Reproducible research
•Path to production
DATA MANAGEMENT
•Security
•Governance
•Standards
•Low maintenance
•Low cost
•Self-service access

THE TYPICAL SOLUTION
“If I can’t use my favorite tools, I’ll…”
• Copy data to my laptop
• Copy data to a data science appliance
• Copy data to a cloud service
Why this is a problem:
• Complicates security
• Breaks data governance
• Adds latency to process
• Makes collaboration more difficult
• Complicates model management and
deployment
• Creates infrastructure silos

CLOUDERA DATA SCIENCE WORKBENCH
Accelerate Machine Learning from Research to Production
•
•
•
•
•

CDSW ARCHITECTURE
Extends traditional clusters with new ML capabilities
• Built with Docker and Kubernetes
• Isolated, reproducible user environments
• Supports both big and small data
• Local Python, R, Scala runtimes
• Schedule & share GPU resources
• Scale to CDH/HDP with Spark, Impala, Hive
• Secure and governed by default
• Easy, audited access to Kerberized clusters
• Leverages shared platform services
• Deployed with Cloudera Manager or
package install (Ambari)
CDH/HDP CDH/HDP
Cloudera Manager/Ambari
gateway node(s) CDH nodes
Hive/Impala, HDFS,
...
CDSW CDSW
...
Master
...
Engine
EngineEngine
EngineEngine
Tristan

ACCELERATED DEEP LEARNING WITH GPUS
Multi-tenant GPU support on-premises or cloud
• Extend CDSW to deep learning
• Train on GPUs, deploy on CPUs
• Works on-premises or cloud
CDSW
GPUCPU
CDH/HDP
CPU
CDH/HDP
single-node
training
distributed
training, scoring
“Our data scientists want GPUs, but
we need multi-tenancy. If they go to
the cloud on their own, it’s expensive
and we lose governance.”
GPU CPU GPU

A MODERN DATA SCIENCE ARCHITECTURE
Containerized environments with scalable, on-demand compute
• Built with Docker and Kubernetes
• Isolated, reproducible user environments
• Supports both big and small data
• Local Python, R, Scala runtimes
• Run Spark, Impala, and other CDH services
• Secure and governed by default
• Easy, audited access to Kerberized clusters
• Leverages SDX platform services
• Deployed with Cloudera Manager
CDH CDH
Cloudera Manager
gateway node(s) CDH nodes
Hive, HDFS, ...
CDSW CDSW
...
Master
...
Engine
EngineEngine
EngineEngine

ACCELERATED DEEP LEARNING WITH GPUS
Multi-tenant GPU support on-premises or cloud
• Extend CDSW to deep learning
• Train on GPUs, deploy on CPUs
• Works on-premises or cloud
CDSW
GPUCPU
CDH
CPU
CDH
CPU
single-node
training
distributed
training, scoring
“Our data scientists want GPUs, but
we need multi-tenancy. If they go to
the cloud on their own, it’s expensive
and we lose governance.”
GPU On CDH coming in C6

Confidential-Restricted – For Discussion Purposes Only
HDP Edge
Node
HDP
Node
HDP
Node
HDP
Node
Ambari
CDSW
Worker Node
HDFS, Hive, HBase, Spark, Phoenix…
HDP Edge Node
CDSW Master Node
Browser
HDP Edge
Node
CDSW
Worker Node
Cloudera Data Science Workbench Nodes
CDSW on HDP Architecture

Confidential-Restricted – For Discussion Purposes Only
CDSW 1.5.0 Support Matrix
● CDH 5
● CDH 6
● HDP 2.6.5
● HDP 3.1.0

DATA CATALOG
GOVERNANCESECURITY LIFECYCLEWORKLOAD XM
STORAGE Amazon
S3
Microsof
t ADLS
HDFS KUDU
INTRODUCING CLOUDERA MACHINE LEARNING
Cloud-native enterprise machine learning platform
DATA SCIENCE DATA ENGINEERING MODEL OPERATIONS
CLOUDERA ML RUNTIME
Python/R, Spark, TensorFlow, CPU/GPU-Optimized
Interactive Development Batch Pipelines Predictive APIs
Full capability of CDSW
Rapid cloud provisioning
and elastic autoscaling
Unified data engineering and
ML with seamless
dependency management
Multi-cloud portability
powered by Kubernetes
Connects to HDFS or
cloud object storage
and shared metadata
Accelerated deep learning
with distributed GPU training
* Initially targeted for cloud
managed K8s services, then
OpenShift
KUBERNETES
EKS, AKS, GKE, OpenShift

WHAT DATA SCIENCE TEAMS DO
Ingest data at scale.
Store and secure data.
Clean and transform data
for analysis.
Explore data and build
predictive models, offline.
Evaluate and tune models.
Develop and deliver a
modeling pipeline.
Test, verify, and approve
model for deployment.
Create and maintain
batch/stream pipelines,
embedded models, APIs.
Update models in
production.
PREPARE DATA BUILD MODELS DEPLOY MODELS

NEW: CLOUDERA DATA SCIENCE WORKBENCH 1.5
Accelerate and simplify machine learning from research to production
ANALYZE DATA TRAIN MODELS
•
DEPLOY APIs
•
NEW! NEW!
MANAGE SHARED RESOURCES

INTRODUCING EXPERIMENTS
Versioned model training runs for evaluation and reproducibility
Data scientists can now...
• Create a snapshot of model code,
dependencies, and configuration
necessary to train the model
• Build and execute the training run in an
isolated container
• Track specified model metrics,
performance, and model artifacts
• Inspect, compare, or deploy prior models

INTRODUCING MODELS
Machine learning models as one-click microservices (REST APIs)
score.py
forecast
f = open('model.pk', 'rb')
model = pickle.load(f)
def forecast(data):
return model.predict(data)

MODEL MANAGEMENT
View, test, monitor, and update models by team or project

CLOUDERA FAST FORWARD LABS

47
CLOUDERA FAST FORWARD LABS
ADVISING &
RESEARCH
ML APPLICATION
DEVELOPMENT
ML STRATEGY
ENGAGEMENT
ML application
strategy prescription ML expert advising
research reports and
prototypes
Expert guidance to accelerate value and scale

AS NEW TECH
CAPABILITIES EMERGE,
BE READY

Machine Learning in the Enterprise 2019

Recommended

More Related Content

What's hot (20)

Similar to Machine Learning in the Enterprise 2019 (20)

More from Timothy Spann (20)

Recently uploaded (20)

Machine Learning in the Enterprise 2019