NVIDIA GPUs in the Cloud 4 EVOLVING CLOUD REQUIREMENTS On Off - - PowerPoint PPT Presentation

NVIDIA GPUs in the Cloud

4

SOFTLAYER CAPABILITIES OVERVIEW Hybrid Cloud Off premises On premises

EVOLVING CLOUD REQUIREMENTS

Connecting clouds New workloads Components to disrupt

5

GLOBAL CLOUD PLATFORM

Only infrastructure solution with a common management interface and API across a unified architecture. Mix and match bare metal servers, virtual servers and turnkey private clouds, and manage them from a single control pane or API. All deployed on-demand and provisioned automatically in real time.

SOFTLAYER CAPABILITIES OVERVIEW

Unified architecture enabled by powerful software

6

GLOBAL FOOTPRINT

SOFTLAYER CAPABILITIES OVERVIEW

7

DATA CENTER POD DESIGN

Each pod includes:

• 10,000 ft2

isolated zone

• 2 megawatts

(n+1 power)

• 150 racks
• 4,000 physical nodes
• Firewalls,

load balancers, and storage

Server Racks

Power Battery Backup Generators Security Network Storage Environmental Controls Each data center consists of four or more “pods.”

SOFTLAYER CAPABILITIES OVERVIEW

8

TRIPLE-NETWORK ARCHITECTURE

• High-performance public network with transit from multiple tier-1 carriers
• Secure OOB management via VPN
• Private network for intra-application and inter-facility communications, access to shared

services

• Native IPv6 support
• Virtual racks for integrated management
• Complete suite of network services

SOFTLAYER CAPABILITIES OVERVIEW

9

ROBUST, FULL-FEATURED API

Functions include:

• Automatic server

deployment

• Service provisioning
• Reboots and reloads
• Ticketing
• Hardware

configuration

• Software load
• DNS
• Network
• Storage
• Security scans
• Monitoring

SOFTLAYER CAPABILITIES OVERVIEW

• Improves customer

control, reduces error, increases flexibility

• SoftLayer API

supports 3,841 function calls to over 279 services

• Supports REST,

SOAP and XML-RPC interfaces

• Enables full

auto-scaling implementations

• Comprehensive

documentation, libraries, and support

10

HOW IT ALL FITS TOGETHER

SoftLayer Infrastructure Management System

• Bare metal and virtual server provisioning
• Integrated BSS/OSS
• Comprehensive network management

Data Center & Pods

• Standardized, modular hardware configs
• Lower inventory carrying costs
• Maximize asset utilization and profitability
• Increase provisioning flexibility
• Simplify capacity management
• Globally consistent service portfolio

Triple Network

• Proprietary network architecture
• Pod design allows customers grow across multiple

racks or rows in the same layer 2/3 domain as needed. SOFTLAYER CAPABILITIES OVERVIEW

11

SOFTLAYER ADVANTAGES

Complete control, access, and transparency Seamless fault-tolerant, multi-site topography Complete portfolio available

• n-demand in all data centers

Single-tenant and multi-tenant environments SOFTLAYER CAPABILITIES OVERVIEW

14

SOFTLAYER CUSTOMER MOMENTUM

SOFTLAYER CAPABILITIES OVERVIEW

Software as a Service Moving to the Cloud Using Next Gen enterprise applications

15

INTEGRATION WITH IBM

Think it. Build it. Tap into it.

Enabling business transformation Business Process as a Service Marketplace of high- value consumable business applications Software as a Service Composable and integrated application development platform Platform as a Service Enterprise class,

• ptimized

infrastructure Infrastructure as a Service

Big Data & Analytics Smarter Commerce Smarter Workforce GBS Cloud Business Solutions Watson Solutions Software Solutions Smarter Cities BluemixTM Cloud Managed Services Infrastructure Services On Premises Cloud Infrastructure

SOFTLAYER CAPABILITIES OVERVIEW

16

HYBRID CLOUD

SOFTLAYER CAPABILITIES OVERVIEW

Shared Off-Premises Cloud Dedicated On-Premises Cloud Traditional IT Dedicated Off-Premises Cloud

Key Considerations:

• Expertise – Best practices
• Open Integration – Hybrid environments
• Control – Visibility, automation

Choose the right mix for your business

Hybrid represents a key element of our differentiated value

17

HPC VDI

NVIDIA TESLA SOLUTIONS www.nvidia.com/tesla NVIDIA GRID SOLUTIONS www.nvidia.com/vdi

18

Tesla k80

world’s Fastest accelerator for data analytics and scientific computing

Caffe Benchmark: AlexNet training throughput based on 20 iterations, CPU: E5-2697v2 @ 2.70GHz. 64GB System Memory, CentOS 6.2, Peak Perf with GPU Boost on

Maximum Performance

Dynamically Maximize Perf for Every Application

Double the Memory

Designed for Big

Data Apps

24GB

2x Faster 2.9 TF| 4992 Cores | 480 GB/s

0x 5x 10x 15x 20x 25x CPU Tesla K40 Tesla K80

Deep Learning: Caffe

Dual-GPU Accelerator for Max Throughput

19

GPUs for Training & Prediction

Prediction Training

Cloud Datacenter

• r

Colocation

20

Image Detection Face Recognition Gesture Recognition Video Search & Analytics Speech Recognition & Translation Recommendation Engines Indexing & Search

Use Cases Early Adopters

Image Analytics for Creative Cloud Image Classification Speech/Image Recognition Recommendation Hadoop Search Rankings

Talks @ GTC

CUDA for Machine Learning

THE END