NEW LAUNCH! Delivering Powerful Graphics-Intensive Applications from the AWS Cloud

Preview:

Citation preview

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

December 2016

CMP320

NEW LAUNCH!

Delivering Powerful,

Graphics-Intensive Apps

from the AWS CloudTosh Tambe, Supreeth Sheshadri, Amazon Web Services

Judd Kaiser, ANSYS

What to Expect from the Session

• Challenges in deploying graphics-intensive

workloads

• Delivering 3D apps from AWS

• Customer use case – ANSYS

Graphics-intensive Workloads

Traditional Workstation

Design & Engineering

Data & Process (PLM)

Conceptual

Design

Engineering Design (CAD)

Simulation & Analysis

(CAE)

Tooling Design (CAM)

Application Toolchain

Application 1

Application 2

Application

Application ..n

Data Management

Application Toolchain on-premises

Application 1

Application 2

Application

Application ..n

Application Toolchain on-premises

Application 1

Application 2

Application

Application ..n

Application Toolchain on-premises

Application 1

Application 2

Application

Application ..n

Application Toolchain on the cloud

Application 1

Application 2

Application

Application ..n

Application Toolchain on the cloud

Application 1

Application 2

Application

Application ..n

Graphics System

Data Model Frame Display DeviceScene

CPU GPU GPU

Data Heavy Data Heavy Data Light

Graphics System - Remoting

Data Model Frame Display DeviceScene

CPU GPU GPU

Data Heavy Data Heavy Data Light

Generate Stream

G2

Generating Graphics on AWS

GPU-graphics Instance – G2

• Launched in Nov 2013

• Up to 4 NVIDIA GRID K520 GPUs

• Suitable for both graphics and compute workload

• Supports DirectX, OpenGL, CUDA, OpenCL

G2 – Some workloads work well

• Applications requiring full GPU

• Applications that consume all of GPU but

require only modest CPU and RAM

• Access hardware encoding

• Use CUDA with OpenGL

G2 – However, other workloads don’t!

G2 has fixed ratio of CPU, RAM, and GPU, which means:

• Either constrained

• Or overprovisioned

ANSYS enables customers to

predict with confidence that their

products will thrive in the real world.

About ANSYS

Our most innovative customers are rapidly deploying simulation

2005 2015 2020

1 in 22 1 in 6 ALL

The ANSYS Vision: Every Engineer Will Use Simulation

ANSYS Enterprise Cloud(an AWS-based simulation data center)

Key solution components

• Scaling HPC (Cycle)

• Interactive 3D graphics (NICE)

• Cloud storage

Enterprise Cloud Graphics

GPU is critical for 3D graphics performance

Need the right balance of GPU/CPU/RAM

Nice DCV “external rendering” feature is

mission-critical

Looking ahead…

Proxy

Op

en

GL

Thin Client

DC

V p

roto

co

l

(htt

ps) G2

R3

Generating Graphics on AWS

Elastic GPU

EC2 + Elastic GPU = Flexibility

t2

c4

m4

r4

:

:

Small GPU

:

:

Large GPU

Attach Elastic GPU to an instance at launch, similar to attaching an EBS volume

Elastic GPU Architecture

Compute + Graphics

Instructions

Gra

ph

ics

Instru

ctio

ns

Instructions

Image

Image

Instance Graphics Attachment

Elastic GPU – Key Features

Range of instance types depending on customer’s need

Graphics attachment in any of a range of sizes

depending on customer’s need

Single tenancy dedicated hardware for the specific

customer and instance

How do you use Elastic GPU?

Console

API / CLI

Managed services

Attaching Elastic GPU – Console

Attaching Elastic GPU – Console

Attaching Elastic GPU – API / CLI

• New parameter added to EC2 ‘RunInstances’ call:

• ElasticGpuType

• CLI Example:

aws ec2 run-instances --ami-id ami123abc --elastic-gpu-type eg1.xlarge

Application Streaming

Our learnings from AppStream

• AppStream – SDK-based streaming engine

• Several gaps – app lifecycle, VPC access,

price/performance, DIY clients

• Customers wanted fully managed streaming platform

Fully managed application streaming service that provides

users instant access to their desktop applications

Desktop Application Streaming

Stream desktop applications securely

to any web browserPay-as-you-go Scale globally

Secure apps & dataRun Desktop Apps

in a Web Browser

Move desktop applications

to the cloud

Provide instant access

to apps from anywhereOffer tools to simplify

application streaming

Why Did We Build Amazon AppStream 2.0?

Business & Public Sector

Move desktop apps

to cloud with no re-write

ISVs Design & Engineering

Benefits: Instant-on Access to Desktop Apps

Instant start for desktop apps – like watching a video online

Benefits: Import Your Apps Without Re-writes

Import existing apps with no changes or re-write and start streaming

Benefits: Works With Your IT

Integrates with existing apps, identity, entitlements, and back-end

Benefits: Fully-Managed Service

No hardware or software to install, submit your apps and start streaming

One streaming instance per end user – no shared instances

Benefits: Consistent Scalable Performance

• Multiple instance types

• Graphics and non-graphics instance families

• Standard, Compute, Memory, and Graphics

• Non-graphics starting from 10 cents/hr

• Supports Elastic GPU

Features: Multiple Instance Types

• Pay per hour for running instances in your fleet

• Scaling policies and instance type choice optimize

cost

• Pay per unique user that connects in a month

• User fee waived for BYOL RDS CALs

AppStream 2.0 Pricing

Build your own streaming SaaS – Elements

Admin UsersStreaming

Technology

• Use multiple apps at the same time

• Clipboard, file upload/download, printing

• Audio and bandwidth controls

• Multiple storage options

• HTML5 browsers with no plug-ins

Simple User Experience

Simple user experience

Simple user experience

NICE DCV streaming protocol

• High fidelity visualization delivered to browsers

• HTTPS access via streaming gateways

• Adaptive and responsive streaming

• AES-256 encrypted

• Supports both 3D and non-graphics applications

Admin set up

1 2 3 4

Network config

On-premises

Public Internet

VPN

or

Direct Connect

Pixels - HTTPS

Identity/SAMLPixels - HTTPS

Streaming GatewayFleet

Utility/License/Database servers

AppStream 2.0 Network – 198.19.x

Customer/ISV VPC

172.X or 192.x or 10.x

Photon Built in

Storage

Private Network Access

HPC Cluster

Stack

Admin Experience

AWS Management Console Programmatic access – AWS SDK

Launch streaming application(s)

$> aws appstream create-streaming-url <stack> [app]

Demo with Elastic GPU

Elastic GPU, AppStream 2.0, and your

applications

• Enable license mobility

• Certify your applications

• Start trials, training, and SaaS environments

Thank you!

CMP321: Building SaaS offerings for desktop apps with Amazon AppStream 2.0

BAP204: Delivering desktop applications to any device anywhere with Amazon AppStream 2.0

Remember to complete

your evaluations!

Recommended