En oversikt, likheter og forskjeller€¦ · En oversikt, likheter og forskjeller Rune Zakariassen Microsoft Historic Computing Transformations We are all excited about the cloudWe

En oversikt, likheter og forskjellerEn oversikt, likheter og forskjeller

Rune ZakariassenMicrosoftMicrosoft

Historic Computing Transformations

We are all excited about the cloudWe are all excited about the cloud IDC Sees Cloud Market Maturing QuicklyIDC Sees Cloud Market Maturing Quickly

I 2009 i l $17 billi l d l d• In 2009, approximately $17 billion was spent on cloud-related technologies, hardware and software.

• By 2013, that spending is expected to grow to $45 billion.y , p g p g $• Frank Gens, senior vice president and chief analyst for the IDC,

declare that the chasm has been crossed and the cloud is well on its way to becoming mainstreamway to becoming mainstream.

Kilde: IDChttp://itmanagement.earthweb.com/features/article.php/3870016/IDC-Sees-Cloud-Market-Maturing-Quickly.htm

In House or Hosted ServersIn House or Hosted Servers

Allocated L dAllocated IT-capacities

“Under-supply“ of capacities

Load Forecast

AC

ITY “Waste“ of

capacitiesFixed cost of IT-capacities

IT C

APA

I

Barrier forActual Load

Barrier forinnovations

TIME

Cloud ComputingCloud Computing

LoadAllocated IT capacities

Load Forecast

No “under-supply“

AC

ITY

Reduction of “over-supply“

Possible reduction of IT

CA

PA

Reduction of initial

investments

IT-capacities in case of

reduced load

I

Actual Load

investments

Time

What Is A Cloud Platform?

“software as a service”

“infrastructure as a service”

“platform as a service”

“information as a service”“everything as a service”

Gartner’s View of Clouds

Cloud computing is cheaper when the economic return is high

CloudCo-location

E i

computingservices

Economiesof scale

Traditionaloutsourcing

On-premiseinstallation outsourcinginstallation

Economies of skill

Data Center Evolution

Leased COLO

Quincy Class

Container Class

Generation 4Modular Data Center

Data CenterData CenterDesignDesign

Deployment Deployment Scale UnitScale UnitScale UnitScale Unit

ContainerS

Rack

Server

The Microsoft CloudThe Microsoft Cloud~100 Globally Distributed Data Centers

Quincy, WA Chicago, IL San Antonio, TX Dublin, Ireland Generation 4 DCs

Workload Patterns Optimal For Cloud

Usage

Compu

te

Ti

Average

Inactivity

Period Average UsageCo

mpu

te

TiTime

On & off workloads (e.g. batch job)On & off workloads (e.g. batch job)Over provisioned capacity is wasted Over provisioned capacity is wasted Time to market can be cumbersomeTime to market can be cumbersome

Time

Successful services needs to grow/scale Successful services needs to grow/scale Keeping up w/ growth is big IT challenge Keeping up w/ growth is big IT challenge Complex lead time for deploymentComplex lead time for deploymentTime to market can be cumbersome Time to market can be cumbersome Complex lead time for deploymentComplex lead time for deployment

Compu

te

Average Usage

Compu

te

Average Usage

Time

Average Usage

Unexpected/unplanned peak in demand Unexpected/unplanned peak in demand S dd ik i t fS dd ik i t f

Time

Average Usage

Services with micro seasonality trends Services with micro seasonality trends Sudden spike impacts performance Sudden spike impacts performance Can’t over provision for extreme cases Can’t over provision for extreme cases

Peaks due to periodic increased demandPeaks due to periodic increased demandIT complexity and wasted capacity IT complexity and wasted capacity

Types of Clouds

PrivatePrivate

Types of Clouds

PlatformPlatformI f t tI f t tPrivate(On-Premise)

Private(On-Premise)

Applications

Platform(as a Service)Platform

(as a Service)

Applicationsnage

Infrastructure(as a Service)

Infrastructure(as a Service)

Applications

Runtimes

Applications

Security & Integration

Runtimes

Applications


You

ma

Runtimes

Applications

Security & Integrationman

age

Databases


age Databases


Manage

Databases


You

m

Servers

Virtualization

You

man Servers

Virtualization

ed by vend

Virtualization

Managed

Servers

Storage

Server HW

Storage

Server HW

dor

Storage

Server HW

d by vendo

Networking NetworkingNetworking

or

Azure & Amazon ComparisonAzure & Amazon Comparison

Y A li i Deplo ment

Frameworks

Your Application Deployment

Web ServerDeployment

Operating System

OS Services Provided byWindows Azure

p g y

ProvidedBy

A

Virtualized Instance

AmazonEC2 Hardware

Azure and Google (AppEngine)

Your ApplicationDeployment Deployment

Frameworks

Web Server

OS ServicesProvided byProvided by

Operating System

OS ServicesGoogle

AppEngine

Provided byWindows Azure


HardwareHardware

Azure & SalesForce.com


Frameworks


Web Server

OS S iProvided by

Operating System

OS ServicesProvided bySalesForce.com Provided by

Windows Azure


Hardware

Data Storage

K t t t i bl b d bl kKey concepts account, container, blob and blocks

BlockBlobContainerAccount

IMG001 JPG

Pictures

IMG001.JPG

IMG002.JPG

AccountBlock AAAA

Movies MOV1.AVI Block AAAB

Block AAACBlock AAAC

Semi‐Structured data

Tables contain entities Tables contain entities

Entities contain properties

May be partitioned across May be partitioned across thousands of servers.

Support ACID transactions ppover single entities

Queries over entire table

.NET and REST interfaces

GetMessageGetMessage (Timeout)(Timeout)RemoveMessageRemoveMessage

MsgMsg 11Worker RoleWorker Role

PutMessagePutMessageMsgMsg 11

MsgMsg 22Web RoleWeb Role MsgMsg 22MsgMsg 11

MsgMsg 33

MsgMsg 44 Worker RoleWorker RoleWorker RoleWorker Role

QueueQueue MsgMsg 22

SQL Azure

Database Replicas

Single Database Multiple Replicas

Single PrimaryReplica 1

g y

Replica 2DB

Replica 3

Scenarios vs. Platform Capabilities

Create Create Very Scalable Web

Moderately Scalable Web

Apps

Create Very Scalable Web

Apps

Create Parallel Processing Apps

Apps with Background Processing

Run On‐Premises Apps

Scale‐out web

VM with standard OS

x x x

x

app platform

Scale‐out batch app platform

x x

x x

x

Scale‐out

Relational storage x

x x

x

storage

Blob storage

x

x x

x

Queues x

GoGrid, Mosso, Flexiscale, OthersTypical scenariosTypical scenarios

Create Very Scalable Web Create



Apps




Apps

Scale‐out web

VM with standard OS VMs

VMsapp platform


VMs

Scale‐out

Relational storage

VMs (w/RDBMS) VMs (w/RDBMS)

storage

Blob storage

Queues

Amazon Web ServicesTypical scenariosTypical scenarios




Apps




Apps

Scale‐out web

VM with standard OS EC2 VMs

EC2 VMs EC2 VMs EC2 VMsapp platform


EC2 VMs EC2 VMs

EC2 VMs , Elastic MapReduce

EC2 VMs

EC2 VMs

Scale‐out

Relational storage

EC2 VMs (w/RDBMS)

EC2 VMs (w/RDBMS)

SimpleDB SimpleDBstorage

Blob storage

SimpleDB

Simple Storage Service (S3) S3

SimpleDB

QueuesSimple Queue Service (SQS)

Windows AzureTypical scenariosTypical scenarios




Apps




Apps

Scale‐out web

VM with standard OS

Web role Web role Web roleapp platform


Web role Web role

Worker role

Web role

Worker role

Scale‐out

Relational storage SQL Azure

Tables Tablesstorage

Blob storage

Tables

Blobs Blobs

Tables

Queues Queues

Google AppEngineTypical scenariosTypical scenarios




Apps




Apps

Scale‐out web

VM with standard OS

Java/Python app platform


runtime

Scale‐out

Relational storage

Datastorestorage

Blob storage

Datastore

Queues

Salesforce.com Force.comTypical scenariosTypical scenarios




Apps




Apps

Scale‐out web

VM with standard OS

Force.comapp platform


runtime

Scale‐out

Relational storage

Force.comstorage

Blob storage

storage

Queues

Comparing Cloud PlatformsSummarizing typical scenariosSummarizing typical scenarios

Run On PremisesCreate Very Scalable Web

Create Very Scalable Web Apps with Background


Create Moderately Scalable WebRun On‐Premises

AppsScalable Web

Apps

gProcessing

Processing AppsScalable Web Apps

GoGrid, Mosso, Flexiscale etc xxFlexiscale, etc. xxAmazon Web

Services x x x x x

Windows Azure x x x xGoogle

AppEngine xAppEngine

Salesforce.com Force.com x

© 2008 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS,

IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

Documents

En oversikt, likheter og forskjeller€¦ · En oversikt, likheter og forskjeller Rune Zakariassen Microsoft Historic Computing Transformations We are all excited about the cloudWe