Applied Micro development platform. ZT Systems (ST based) HP Redstone platform. Mitac Dell Copper platform. ARM in Servers

Applied Micro development

ZT Systems (ST based) platforml tf

ZT Systems (ST based)

Dell “Copper”

HP “Redstone”

platform Mitac

ARM in Servers

Dell Copper platform

ARM in Servers

Server Ecosystem Momentum

 2009: Internal ARM trials hosting part of website on ARM technology

 2010: Calxeda raise $48M$

 ^{H2 2011}

 First release of Ubuntu server Linux for ARM

 HP announce Redstone development platform

 HP announce Redstone development platform

 Applied Micro announce ARMv8 based X-Gene

 Oracle announce server C2 JIT / JVM on ARM

 ^{H1 2012}

 ^{H1 2012}

 Ubuntu release 12.04LTS server for ARM

 First public demo of ARM platform with 12.04LTS,

MAAS J J O St k d j R R

MAAS, JuJu, OpenStack, node.js, RoR

 Dell “Copper” announcement

 Mitac GFX Server announcement

 ^{H2 2012}

 ARM announces 64-bit processors

 Linaro forms software group focused on ARM powered servers

ARM Focuses Where the Server IS the Business

 Server infrastructure is primary generator of profit for social

t ki d h ti i

Social Networking

networking and hosting companies

 Server optimization is critical to

 Server optimization is critical to maximize revenue and minimize running costs g



Cloud Hosting

 Businesses designed to be adaptive to new technology

to new technology



One Size Does Not Fit All

 Different server workloads require different blend compute networking and storageDifferent server workloads require different blend compute, networking and storage

 ARM’s semiconductor partners can design specific SoCs for specific end markets

 ARM’s Cortex processor technology ideal for running lower CPU intensive tasks

 Light scale out examples include Static web servers Content delivery (i e video music)

 Light scale-out examples include Static web servers, Content delivery (i.e. video, music),

Large distributed memory caching, Simple search systems (Hadoop, offline data analytics etc)

Servers Will Mimic Mobile SOC Evolution

 Specific workloads for servers driving

segmentation in the

Key Drivers of the Potential Evolution of ARM-based Server Chips

segmentation in the server market

 Generic, one-size-fits-all

CPU l it bl th ^e/$

Linux O/S Multiple Small CPU’s

Integrated Fabric

Heterogeneous Processing SOC’s

H/W Accelerators

CPUs less suitable than specifically designed, fully integrated SoCs

mance/Joule

O/S Multiple Apps

Linux O/S Single Application

Large Core CPU Virtualization

g

Power Efficiency

 Computing efficiency

becoming as important in servers as it is in mobile

Perform Multiple Apps CPU

Cloud

Distributed Computing

 Semiconductor

companies are increasing the compute efficiency of

Time

the compute efficiency of ARM-based server chips by integrating hardware blocks onto the SoC

blocks onto the SoC

32-bit Software Ecosystem in Place

 All critical software pieces now in place to enable initial

ARM l tf t hi

ARM server platforms to ship by year end









 Intense interest/commentary across press, analyst and end p y user communities

 Evaluations of ARM servers

now underway at some tier 1

data centers

32-bit Software Ecosystem in Place

Common frame orks a ailable LAMP stack running Calxeda.com

Linux + Apache + MySQL + PHP

Common frameworks available Java/Tomcat, Ruby on Rails, Python, Perl

OpenStack enabled ARM servers Dynamically provision new

i t di tl t ARM d

Canonical’s Juju provisioning

a cluster of ARM based servers instances directly onto ARM nodes.

a cluster of ARM-based servers

Open Source Server Community



tools



deliver optimized core open-source software for ARM servers



www linaro org/ser

product time to market



www.linaro.org/ser



ver

to focus on innovation and differentiated value-add

32-bit SOCs for Servers

 Marvell ARMADA-XP





 Calxeda EnergyCore









 TI KeyStone II (announced) y ( )





Selected 32-bit ARM Powered Servers

 Dell: “Copper” platform based on Marvell (ARMv7) silicon

 HP: “Redstone” platform using

Calxeda EnergyCore device (Cortex- A9)





 ^{Mit GFX b d M ll}

 Mitac: GFX server based on Marvell

silicon.

ARM Technology for Servers

Cortex

Cortex--A Series A Series

“Low

“Low--Power Leadership”Power Leadership”

Cortex-A57 Cortex A57

Cortex-A53

- Performance,

- Technology leadership

- Best Performance per mW

Cortex-A15

Cortex-A9

Cortex-A7

Cortex-A5 Cortex-A8

Today 2013 Future

Cortex-A57: Performance and Capability

 Maximum performance for ARM based SOCs



 Revolutionizing computing



 Improved privacy and security framework

 Improved privacy and security framework



 Enhanced capabilities for enterprise

 Enhanced capabilities for enterprise







ARMv8 (64-bit) Server SOCs in Development

ARMv8 Architecture Licensees

Cortex-A50 Licensees

Cortex-A50 Licensees

Applied Micro: X-Gene Development







 28nM version (2014) will support 32 cores running up to 3GHz. Coherency between four SoCs

ARM’s Server Opportunity

 “Energy constrained” servers are driving innovative low- power architectures/platforms

 ARM is well placed to take advantage of this transition







 Market size ~12m servers per year



O t it f 4 ARM P d S C t













Summary

 ARM and its partners are poised to take advantage of three disruptive elements







 Initial (32-bit) ARM powered servers will ship into market this year

 64-bit ecosystem development underway to support system hi t i 2014

shipments in 2014