FUT-1230 - Artificial intelligence at the speed of excellence · 2020. 6. 4. · 21 SUSE Artificial Intelligence Stack Version 1.0 Version 2.0 Version 3.0 Version 4.0 NVIDIA - INTEL

1

Artificial Intelligence At The Speed Of Excellence

FUT-1230

Roadmap & Vision

Alessandro Festa

Sr. Product Manager Artificial Intelligence

[email protected]

@bringyourownai

2

Agenda

1. Artificial Intelligence Landscape Challenges

2. Building a solution from SUSE perspective

3. Roadmap Vision

4. Roadmap Strategy

3

What is the real

challenge of

Artificial Intelligence

for Enterprises?

The Commodity Dilemma

4

ai

This is where you want to be.

You are here.

…Things to be done…

5

Building a

solution from

SUSE’s

perspective

6

SUSE AI Playground

I need a simple

way to..

Develop my AI models without

spending time to configure my

environment

Deploy an AI platform for my

Data Scientist without waste

time in learning how to do so..

Monitor what Data Scientist

deployed on the AI platform

without waste time in configure

all I need…

Data Scientist

Ai Operator

OR

7

Risky

Outcomes

More Production

DelaysFalse promises

Unclear outcome,

uncertainty of

infrastructure stability

Production DelaysHigh risk of

errors“customer” walk away

and complain

Ai operator need

to deploy a new Ai

platform Release Ai

platforms

Now what? Am I

sure everything

works as

expected?

Setup/configure

monitoring

How do I ensure everything

works as expected? What I

need to monitor? What I

need to check?

Deploy Ai Platform

How do I install the Ai

Platform? What I need

(storage/containers)?

Define Software

“basics”

What we need as basics?

K8’s/HPC/Single

Server/Cloud?

Define HW resources

What we need?

GPU’s/CPU’s what if we

add/change/remove later?

Collect Data

Scientist

requirements

Data Scientist are not

infrastructure experts.

DevOps are not Ai

experts.

Challenges

Artificial Intelligence Value Stories

“Traditional processes are not good enough any

longer” - An example: Ai Operator journey

Ai playground solve this Ai Stack solve this

8

Ai operator need

to deploy a new Ai

platform

Release Ai

platforms

Now what? Am I

sure everything

works as

expected?

Setup/configure

monitoring

How do I ensure everything

works as expected? What I

need to monitor? What I

need to check?

Deploy Ai Platform

How do I install the Ai

Platform? What I need

(storage/containers)?

Define Software

“basics”

What we need as basics?

K8’s/HPC/Single

Server/Cloud?

Define HW resources

What we need?

GPU’s/CPU’s what if we

add/change/remove later?

Collect Data

Scientist

requirements

Data Scientist are not

infrastructure experts.

DevOps are not Ai

experts.

Challenges

Use a “ready-to-go”

solution like Ai

Playground allow the

AiOps to select from

pre-configured set of

templates

The pre-built

automation workflow

remove complexity

and give freedom to

change the

infrastructure even

at later time

Templates are

organized by

“target”: on-prem,

on-cloud, dev-

sandboxes and by Ai

platform

We drive the AiOps

through the entire

deployment

experience. Just

launch the auto-pilot.

We offer a tested and

maintained solution.

We take care also of

the “sides” not just

of the “main course”

Platform is ready.

We give you the

basic instructions to

pass to your users.

Job is done.

Mitigation

Artificial Intelligence Value Stories

Real

Outcomes

No delays makes

people happyWe may offer a

solution

We do reduce frictions

and accelerate projects

We stick with deliver

times

Deploy what

you need and

nothing more..

“customer” is happy

and ask for more

9

Roadmap Vision

10

Building an AI

Playground

Avoiding Complexity for Data Scientist

11

GPU/CPU

SLES/openSUSE

MacOS

Container Runtime

SUSE DSVM

AI Playground

Jupyter MS VsCode Nteract Atom

OS Selection

Alternative Component

AI Editor/Ai

Template of choice

HW Selection

Windows WSL2

PolyNote R-StudioML/DL

FrameworksAI Remote

(Pipelines)

SUSE Artificial Intelligence Playground- Reference Architecture

12

Version 1.0 Version 2.0 Version 3.0 Version 4.0

NVIDIA - INTEL

SLE(15 or above)

CAASP or K8’s (including

Cloud Providers)

KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

NVIDIA-INTEL-AMD

SLE(15 or above)


Cloud Providers)

HPC (on prem-on cloud)

KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

NVIDIA-INTEL-AMD-HUAWEI

SLE(15 or above)


Cloud Providers)


CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

NVIDIA-INTEL-AMD-

HUAWEI-…

SLE(15 or above)


Cloud Providers)


CAP

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

x86 X86 - ARM X86 - ARM X86 - ARM

JUPYTER

HUB

KUBEFLOW

–

AIRFLOW

JUPYTER

HUB

KUBEFLOW

AIRFLOW

…

JUPYTER

HUB

Ai Containers

Ai Platforms

Ai Application

Orchestrators

Ai Infrastructure

Orchestrators

Ai Storage

Ai OS

HW Accelerators

Platforms

3 months 3 months 3 months

SUSE Artificial Intelligence Stack- Reference Architecture

13

SUSE Artificial Intelligence PlaygroundReference Architecture

GPU/CPU

SLES/openSUSE

MacOS

Container Runtime

SUSE DSVM

AI Playground (Electron)

Jupyter MS VsCode Nteract Atom

OS Selection


AI Editor of choice

HW Selection

Windows WSL2

PolyNote R-StudioML/DL

FrameworksAI Remote

(Pipelines)

14

SUSE Artificial Intelligence Playground

OS

Selection

Alternative

Component

AI Editor

of choice

HW

Selection GPU/CPU

SLES/openSUSE

Container Runtime

AI Playground (Stratos)

Version 1

Single Container CPU Single Container GPU build

at runtime

Single VM CPU (when DSVM will be ready)

Jupyter NotebookJupyter LabAtomMicrosoft VS CodeNteract

Ai Packages

15


OS

Selection

Alternative

Component

AI Editor

of choice

HW

Selection GPU/CPU

SLES/openSUSE

Container Runtime


Version 1

Single Container CPU Single Container GPU build

at runtime

Single VM CPU (when DSVM will be ready)

TerrafornSkubaKfctlKubectl

Helm

Infra Deployment

Ai Platform Deployment

Ai Infra Monitoring

16


OS

Selection

Alternative

Component

AI Editor

of choice

HW

Selection GPU/CPU

SLES/openSUSE

Container Runtime


GPU/CPU

SLES/openSUSE

Container Runtime

SUSE DSVM


Windows WSL2

Version 1 Version 2

17


OS

Selection

Alternative

Component

AI Editor

of choice

HW

Selection GPU/CPU

SLES/openSUSE

Container Runtime


GPU/CPU

SLES/openSUSE

Container Runtime

SUSE DSVM


Windows WSL2

GPU/CPU

SLES/openSUSE

MacOS

Container Runtime

SUSE DSVM


Windows WSL2

Version 1 Version 2 Version 3

18

AI Software

Stacks

Infrastructure Ready To Go For Your AI Journey

19

SUSE Artificial Intelligence Stack- Reference Architecture

GPU/CPU

HOST (for K8’s/HPC)

CAASP or K8’s (including Cloud Providers)


CAP

AI Platform

CONTAINER CONTAINER VM VM

Optional Component


AI Platform of choice

GPU/CPU enabled containers

(including third-party registries)

OS optimized for HW

SES or CephStorage

SLE(15 or above)

20

SUSE Artificial Intelligence StackVersion 1.0 Version 2.0 Version 3.0 Version 4.0

NVIDIA - INTEL

SLE(15 or above)


Cloud Providers)

KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

NVIDIA-INTEL-AMD

SLE(15 or above)


Cloud Providers)


KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER


SLE(15 or above)


Cloud Providers)


CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

NVIDIA-INTEL-AMD-

HUAWEI-…

SLE(15 or above)


Cloud Providers)


CAP

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

x86 X86 - ARM X86 - ARM X86 - ARM

JUPYTER

HUB

KUBEFLOW

–

AIRFLOW

JUPYTER

HUB

KUBEFLOW

AIRFLOW

…

JUPYTER

HUB

Ai Containers

Ai Platforms

Ai Application

Orchestrators

Ai Infrastructure

Orchestrators

Ai Storage

Ai OS

HW Accelerators

Platforms

AWS “How-to” Documentation (NVIDIA)

SuperMicro Reference Architecture (NVIDIA/Intel)

Intel “Stacks” as “SUSE-Intel Stacks”

AMD Reference Architecture

ARM Reference Architecture (not CaaSP, include NVIDIA)

Huawei Reference Architecture (Huawei GPU’s)

Dell Reference Architecture (NVIDIA/Intel/AMD)

Reference Architectures and Joint “go-to-market”

21

SUSE Artificial Intelligence StackVersion 1.0 Version 2.0 Version 3.0 Version 4.0

NVIDIA - INTEL

SLE(15 or above)


Cloud Providers)

KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

NVIDIA-INTEL-AMD

SLE(15 or above)


Cloud Providers)


KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER


SLE(15 or above)


Cloud Providers)


CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

NVIDIA-INTEL-AMD-

HUAWEI-…

SLE(15 or above)


Cloud Providers)


CAP

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

x86 X86 - ARM X86 - ARM X86 - ARM

JUPYTER

HUB

KUBEFLOW

–

AIRFLOW

JUPYTER

HUB

KUBEFLOW

AIRFLOW

…

JUPYTER

HUB

Ai Containers

Ai Platforms

Ai Application

Orchestrators

Ai Infrastructure

Orchestrators

Ai Storage

Ai OS

HW Accelerators

Platforms

Manifests

Skuba

Terraform

Reference Architectures Roadmap Vision

AWS

BareMetal

Kustomize

…

SUSE Custom Manifest

AWS

…

22

SUSE Artificial Intelligence Stack- Reference ArchitectureVersion 1.0 Version 2.0 Version 3.0 Version 4.0

NVIDIA - INTEL

SLE(15 or above)


Cloud Providers)

KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

NVIDIA-INTEL-AMD

SLE(15 or above)


Cloud Providers)


KUBEFLOW

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER


SLE(15 or above)


Cloud Providers)


CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

NVIDIA-INTEL-AMD-

HUAWEI-…

SLE(15 or above)


Cloud Providers)


CAP

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

CON

TAIN

ER

SES or Ceph

x86 X86 - ARM X86 - ARM X86 - ARM

JUPYTER

HUB

KUBEFLOW

–

AIRFLOW

JUPYTER

HUB

KUBEFLOW

AIRFLOW

…

JUPYTER

HUB

Ai Containers

Ai Platforms

Ai Application

Orchestrators

Ai Infrastructure

Orchestrators

Ai Storage

Ai OS

HW Accelerators

Platforms

3 months 3 months 3 months

23

Roadmap Strategy

24

Vision

From zero to the top

The need of the «less (effort) is better»

Be different

25

AI Packages

26

AI Virtual Machines For Data Scientist

openSUSE Leap 15.2

https://download.opensuse.org/repositories/Virtualization:/Appliances:/Images:/openSUSE-Leap-15.2/images/

Containers may be created out of every

single package.

https://download.opensuse.org/repositories/Virtualization:/Appliances:/Images:/openSUSE-Leap-15.2/images/

27

What Are AI “Packs”- AI Playground

Any OS :

As application on Data Scientist environment:

• SUSE Linux

• Windows (WSL)

• Web (Jupyter)

Any Tool

As «known» interface to model locally:

• Jupyter Lab

• Jupyter Notebook

• Jupyter Nteract

• Polynote

• MS VSCode

• Atom

• …

Any Framework

As container to be deployed everywhere and to be rapidly accessible locally or remote:

• Tensorflow

• PyTorch

• OpenCV

• …

Any Deployment

Once model is ready is deployable as pipeline in a remote enviroment:

• SUSE CaaSP

• AKS (Azure Arc)

• EKS

• GKS

• …

28

April 2020 October 2020 June 2021 2022

SUSE Artificial Intelligence1.0

1.1

2.0

Ai Playground• Editors support:

• Jupyter

• JupyterLab

• MS VsCode

• Nteract

• Templates:

• Kubeflow

• Monitoring:

• Kubeflow

AI “Stacks”• Kubeflow on CaaSP

• Intel-SUSE Stacks

• Supermicro Reference Architecture

• Dell Reference Architecture

• ARM Reference Architecture

AI packages• Tensorflow 1.15

• Tensorflow 2.1

• Pytorch

• Caffe

• OpenCV

• ONNX

• ArmNN

• R-Studio

1.0

Ai Playground• Editors support:

• Atom

• Polynote

• R-Studio

• Apache Zeppelin

• Templates:

• AirFlow

• Monitoring:

• Airflow

• Deployment

• Kubeflow

AI “Stacks”• AMD-SUSE Stacks

• NVIDIA Reference Architecture

AI packages• Apache Beam

• Apache Spark

• Tensorflow TFX

Ai Playground• Templates:

• Apache Spark

AI “Plugins”

• IdM/PAM plugins

• Authentication/Authorization

plugins

Ai Playground• Templates:

• AirFlow

• MLRun

• Apache Spark

• Monitoring:

• Apache Spark

• Deployment

• Airflow

• MLRun

• Apache Spark

1.1 2.0 2.1

* Information is forward looking and subject to change at any time.

Jan 2021

2.1

** Items are tech preview

29

General Disclaimer

This document is not to be construed as a promise by any participating company to

develop, deliver, or market a product. It is not a commitment to deliver any material,

code, or functionality, and should not be relied upon in making purchasing

decisions. SUSE makes no representations or warranties with respect to the contents of

this document, and specifically disclaims any express or implied warranties of

merchantability or fitness for any particular purpose. The development, release, and

timing of features or functionality described for SUSE products remains at the sole

discretion of SUSE. Further, SUSE reserves the right to revise this document and to

make changes to its content, at any time, without obligation to notify any person or entity

of such revisions or changes. All SUSE marks referenced in this presentation are

trademarks or registered trademarks of SUSE, LLC, Inc. in the United States and other

countries. All third-party trademarks are the property of their respective owners.

FUT-1230 - Artificial intelligence at the speed of excellence · 2020. 6. 4. · 21 SUSE Artificial Intelligence Stack Version 1.0 Version 2.0 Version 3.0 Version 4.0 NVIDIA - INTEL

Documents