DevOps for Software and Systems - MathWorks

0

DevOps for Software and Systems:

Putting Algorithms and Models in Operation

Peter Brady Martin Becker

22

Meaghan Kosmatka

Senior Engineer at Deere & Company

Arvind Hosagrahara

Chief Solutions Architect at MathWorks

33

John Deere saved costs by using MATLAB and Simulink to model

damage calculations in the cloud

44

Companies are transforming their businesses

Selling physical

assets

Selling products

with fixed

functionality

Revenue from

one-off sales

Selling uptime,

logistics support

Selling products

that are updated

in the field

Subscription

based pricing

models

55


Selling physical

assets

Selling products

with fixed

functionality

Revenue from

one-off sales

Selling uptime,

logistics support

Selling products

that are updated

in the field

Subscription

based pricing

models

Digital Products and Services

• Fleet Management• Predictive Maintenance• Smart Machine Optimization• Anomaly Detection

66


Selling physical

assets

Selling products

with fixed

functionality

Revenue from

one-off sales

Selling uptime,

logistics support

Selling products

that are updated

in the field

Subscription

based pricing

models

Digital Products and Services

• Fleet Management• Predictive Maintenance• Smart Machine Optimization• Anomaly Detection

77Source: Gartner: “Use Gartner’s 3-Stage MLOps Framework to Successfully Operationalize Machine Learning Projects”, Shubhangi Vashisth, Gartner ID: G00725627; July 2, 2020

Source: Gartner “Innovation Insight for ModelOps”; Farhan Choudhary; Gartner ID G00729419; August 6, 2020.

“Approximately half of all AI models never make it into production due to lack of ModelOps”

88Icons made by Freepik from www.flaticon.com

99

DevOps“A set of practices and culture that combine software development (Dev) and IT operations (Ops)”

1010


Model“A MATLAB algorithm, or a mathematical, machine learning, deep learning, or Simulink model”

1111


Model +

“Set of best practices, workflows, and tools that combine model development and model operations”

“A MATLAB algorithm, or a mathematical, machine learning, deep learning, or Simulink model”

DevOpsModel

1212

Realize the business value of your models by applying DevOps

processes and deploying them into production

DevOpsModel

Development Operations

DESIGN

BUILD

TEST

DEPLOY

OPERATE

MONITOR

1313

Model DevOps brings agile processes to both data-driven and

physics-based models

DevOpsModel


Data-driven:

• Machine Learning

• Deep Learning

Physics-based:

• Mechanical

• Electronic

• Hydraulic

Code-based:

• Optimization

• Financial

DESIGN

BUILD

TEST

DEPLOY

MONITOR

OPERATE

1414

Model DevOps brings agile processes to both data-driven and

physics-based models

DevOpsModel


• Docker containers

• Software component

• Edge systems

• Embedded devices

Data-driven:

• Machine Learning

• Deep Learning

Physics-based:

• Mechanical

• Electronic

• Hydraulic

Code-based:

• Optimization

• Financial

DESIGN

BUILD

TEST

DEPLOY

MONITOR

OPERATE

1515

Getting access to data on-premises and in the cloud

Integrating with design, build, test, and development toolchains

What are some of the challenges of working with models?

Collaborating with other departments, for example IT, data scientists, production operations

1616

MATLAB and Simulink work with your organization’s development

platforms and toolchain


DESIGN

BUILD

TEST

DEPLOY

MONITOR

OPERATE

1717

DESIGN

BUILD

TEST

DEPLOY

MONITOR




Data science platform

OPERATE

1818

DESIGN

BUILD

TEST

DEPLOY

MONITOR





CI/CD

Source Control

OPERATE

1919

BUILD

DESIGN

TEST

DEPLOY

MONITOR


Operationalize your models on embedded, edge, or IT systems

without recoding

CI/CD


Source Control

Link: Continuous Integration plugins

OPERATE

https://www.mathworks.com/help/matlab/matlab_prog/continuous-integration-with-matlab-on-ci-platforms.html

2020

BUILD

DESIGN

TEST

DEPLOY

MONITOR



without recoding

IT/OT Systems

{ RESTful API }

CI/CD


Source Control


OPERATE


2121

BUILD

DESIGN

TEST

DEPLOY

MONITOR



without recoding

Edge

IT/OT Systems

Embedded

{ RESTful API }

CI/CD


Source Control


x86C/C++

ARM

FPGAC/C++

SOC/PLC

OPERATE


2222

BUILD

DESIGN

TEST

DEPLOY

MONITOR



without recoding

Edge

IT/OT Systems

Embedded

{ RESTful API }

Logging / Dashboard

CI/CD


Source Control


x86C/C++

ARM

FPGAC/C++

SOC/PLC

OPERATE


2323

BUILD

TEST

DEPLOY

MONITOR


Model Based Design maps nicely onto the Model DevOps workflow

Deployment

MBSE ↔MBD

System Simulation

Edge

IT/OT Systems

Embedded

{ RESTful API }

x86C/C++

ARM

FPGAC/C++

SOC/PLC

BUILD

DESIGN

Code Generation

Model- and Code-based V&V

OPERATE

2424

DESIGN

BUILD

TEST

DEPLOY

MONITOR

Data is essential for both development and operations

Da

ta S

ou

rce

s • JDBC/ODBC

• AWS S3, Azure BLOB

• CosmosDB, Athena, DynamoDB

• Azure Data LakeDatabases

Files


MathWorks GitHub repository: https://github.com/mathworks-ref-arch

Pre-processing

Labeling

TrainingOPERATE

https://github.com/mathworks-ref-arch/matlab-production-server-on-aws

2525

DESIGN

BUILD

TEST

DEPLOY

MONITOR

Data is essential for both development and operationsB

ig D

ata

An

aly

tics

cloudera

databricks

ApacheSpark

hadoopD

ata

So

urc

es • JDBC/ODBC




Files



Pre-processing

Labeling

TrainingOPERATE


2626

DESIGN

BUILD

TEST

DEPLOY

MONITOR

Data is essential for both development and operationsB

ig D

ata

An

aly

tics

cloudera

databricks

ApacheSpark

hadoopD

ata

So

urc

es • JDBC/ODBC




Files

Str

ea

min

g

Da

ta

Op

era

tio

na

l

Tech

no

log

y



Da

sh

bo

ard

s

Pre-processing

Labeling

Training

Ingestion

Visualization

OPERATE


2727

DESIGN

BUILD

TEST

DEPLOY

MONITOR

Diverse teams across Engineering, IT, Line of Business, and

Operations must collaborate to achieve success in Model DevOps

DevOps Engineer

Controls Engineer

Systems Engineer

Analyst

Embedded Engineer

IT Software Engineer

Icons made by Freepik from www.flaticon.com

Industrial Engineer

Process Engineer

Data Scientist

IT Architect


OPERATE

2828

Summary

OPERATE

DEPLOY

MONITORTEST

DESIGN

BUILD


Engineering, Data Science, IT, and Operations teams must collaborate to ensure success

MATLAB & Simulink can be integrated into your development environment and leverage data from a variety of data sources

MATLAB & Simulink models can be deployed into a variety of platforms: embedded, edge, IT/OT, and cloud

Model DevOps

29

is a Leader in the 2021 Gartner

Magic Quadrant for Data Science

and Machine Learning Platforms for

the Second Year in a Row

Gartner Magic Quadrant for Data Science and Machine Learning Platforms, Peter Krensky, Carlie Idoine, Erick Brethenoux, Pieter den Hamer, Farhan Choudhary, Afraz Jaffri, Shubhangi Vashisth,1st March 2021.

This graphic was published by Gartner, Inc. as part of a larger research document and should be evaluated in the context of the entire document. The Gartner document is available upon request from MathWorks.

Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research

publications consist of the opinions of Gartner research organization and should not be construed as statements of fact. Gartner disclaims all warranties, express or implied, with respect to this research, including any

warranties of merchantability or fitness for a particular purpose.

3030

? How can we help you apply Model DevOps into your organization?

31

© 2021 The MathWorks, Inc. MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See mathworks.com/trademarks

for a list of additional trademarks. Other product or brand names may be trademarks or registered trademarks of their respective holders.

Thank you

DevOps for Software and Systems - MathWorks

Documents

Transcript of DevOps for Software and Systems - MathWorks