You are on page 1of 47

13.03.

2009

Beat Schwegler
beatsch@microsoft.com
Enterprise & Technical Evangelism Lead
Microsoft Western Europe

13.03.2009

Increase market share and revenue


Investing in product development and customer
facing interaction channels

Increase efficiency and lower TCO


Investing in technologies and processes to drive
efficiency and lower cost through optimization

13.03.2009

Socialization of IT
Datacenters and Green IT

Utility Computing and IT as a Service

Internet applications that explicitly


leverage network effects

13.03.2009

Unused
capacity

Buy more of
everything
Rearchitect
Database to
scale out

Users
Buy bigger
SQL Box

Servers

Add some
Web Servers

13.03.2009

100%

Watts (% of maximum)

90%

80%

70%

60%

50%

40%
0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Workload (% of maximum throughput)

Windows Server 2003

Windows Server 2008

.achieved power savings approx 10 percent or more over Windows


Server 2003 OOB at comparable levels of throughput.

13.03.2009

Server power is small part of overall bill


CPU $$ is less than 1% of total
Trend is towards more efficient CPU,
Lower rack power densities (amps/foot3) on horizon

Disk is more complicated


Mass storage consumes more than CPU, Fan, Memory,
Face drives combined
Keeping DB on line all the time consumes more than
all CPUs involved

The big power spend is cooling

13.03.2009

13.03.2009

Were running scale services now

2 B Live Search queries/month


10 B MSN page views/month
30 B Live ID authentications/month
240 B Messenger messages/month

Were building to run your services

$500 million per data center


10,000 servers/month
$1 B spend/FY08

Volume

Volume

13.03.2009

$10,000
machine
$500
machine
$500
machine

$1000
machine
$500
machine

Scale Up

$500
machine
$500
machine

Scale Out

# Machines

13.03.2009

Business logic

Expand to new locale


Perform live upgrade for new feature
Apply OS patches
Diagnose service failures
Add storage capacity
Handle increase in traffic
Respond to hardware failures

Service glue
and operations

Datacenter

Cost: $500 million each


Size: 500,000 square foot facility (10 football fields) on average
Container-based

10

13.03.2009

The same facilities as a desktop OS, but on


a set of connected servers:

Abstract execution environment


Shared file system
Resource allocation
Programming environments

And more: Utility computing

24/7 operation
Pay for what you use
Simpler, transparent administration

Impacts CAPEX/OPEX spending


May require SOA in the large

Governance
Identity
Business Processes

New Business Model Emerge


Introduces new Programming Models

11

13.03.2009

Self-Hosted Service Provider


Datacenter
Assets

Software Vendor
Product
Assets
IT Assets
Assets

Product
Assets

Datacenter
Operations

IT Assets

IT Operations

Assets

Expenses

IT Operations
Expenses

Utility Compute Consumer


Product
Assets

Service Usage
Expenses

IT Assets

IT Operations

Assets

Expenses

Utility Compute Consumer

Enterprise Customer
Solution
Assets

Amortizations

IT Assets

IT Operations

Assets

Service Usage

Solution
Assets

Amortizations

IT Assets

IT Operations

Assets

Expenses

Pure SaaS Consumer

Expenses
Service Usage
Expenses

IT Assets

IT Operations

Assets

Expenses

12

13.03.2009

Customer
Service
Provider

Customer 1

Models are based on the book Place to Space by Peter Weill & Michael R. Vitale

Product flow only if hosted


through partner
$

Service
Provider

Customer
P

P
P

Service
Extending
Partner

Service extensions developed by


a partner (revenue sharing with
partner)

Customer

Custom
Developing
Partner

pays for additional value

No payment if hosted at service


provider

Models are based on the book Place to Space by Peter Weill & Michael R. Vitale

13

13.03.2009

Service
Provider

Customer

Customer
Service
Extending
Partner

Custom
Developing
Partner

Models are based on the book Place to Space by Peter Weill & Michael R. Vitale

Eric Brewer's CAP Theorem


Consistency
Availability
Tolerance to network Partitions
One can only achieve a combination of two
This leads to BASE semantic (vs. ACID)
Basically Available
Soft-state
Eventual consistency

14

13.03.2009

Consistency Levels

Consistency Level

Changes are Visible

Example

Strong

Now

Missile Launch

Eventual

In the Future

Address Change

Optimistic

Maybe in the Future

Stock Ticker

Message Assurances

Assurance

Message Delivery

Example

Exactly Once

No loss, no duplicates

Bank Transfer

At Least Once

No loss, duplicates

Email

At Most Once

Loss, no duplicates

Streaming Video

Best Effort

Loss, duplicates

Stock Ticker

15

13.03.2009

Where did we start?

Where did we end up?

Shared State

Partitioned, Replicated State

ACID Transactions

Eventual Consistency

Exactly Once Messaging

Best Effort Messaging

Machine Loss is a Catastrophe

Machine Loss is Business As Usual

Keep Processes Running

Recovery-Oriented Computing

Azure Services Platform

16

13.03.2009

Windows Azure is the foundation


of Microsofts Cloud Platform
It is an Operating System for the Cloud
and provides Essential Services for the
Cloud

Virtualized Computation
Scalable Storage
Automatic Management
Developer SDK

Lots of racks, lots of servers, lots of wire.

17

13.03.2009

8+ cores per server


1,2,4 cores per VM

Lights out model

VM reimaged on failure
Server retired on failure

Quick deployment

100 VMs in ~20 minutes


Go to System Up

Management & Deployment Fabric

18

13.03.2009

WebSite
Site
Web
(ASPX, ASMX,
Web
(ASPX,Role
ASMX,
WCF)
(ASPX,
WCF)
WCF)

L
B

Worker
Worker
Service
Role
Internet

Your Service

Queue
L
B
Blobs

Tables

Storage

Windows Azure Datacenter

19

13.03.2009

Web farm that handles request from the Internet


IIS7 hosted web core

Hosts ASP.NET
Public Internet

Managed code only


Supports SSL
Windows Azure specific CAS policy

Web Role
Load
Balancer

Storage Services

No inbound network connections


Can read requests from queue
in storage
Public Internet
Managed code only
Windows Azure specific CAS policy
Worker
Role

Storage Service

20

13.03.2009

Tech Preview offers one type of VM

Platform: 64-bit Windows Server 2008


CPU: 1.5-1.7 GHz x64 equivalent
Memory: 1.7 GB
Network: 100 Mbps
Transient local storage: 250 GB
Azure storage also available: 50 GB

Full service model will support more options

Public Internet

Web Role
Load
Balancer

Worker
Role

Storage Service

21

13.03.2009

Public Internet

Web Role
Load
Balancer

Worker
Role

Storage Service

Public Internet

Web Role
Load
Balancer

Worker
Role

Storage Service

22

13.03.2009

Public Internet

Web Role
Load
Balancer

Worker
Role

Storage Service

3rd Party
Service

Public Internet
Load
Balancer

Web Role
3rd Party
Service

Worker
Role

3rd Party
Service

Storage Service

23

13.03.2009

Blobs Provide a simple interface for storing


named files along with metadata for the file

Tables Provide structured storage.


A Table is a set of entities, which
contain a set of properties

Queues Provide reliable storage and


delivery of messages for an application

Azure Queues
RemoveMessage
GetMessage
(Timeout)
HTTP/1.1 200 OK
Transfer-Encoding: chunked
Content-Type: application/xml
POST
http://myaccount.queue.core.windows.net/myqueue/messages
PutMessage
Date:
Tue, 09 Dec 2008 21:04:30 GMT
Worker Role
Server: Nephos Queue Service Version 1.0 Microsoft-HTTPAPI/2.0

DELETE
Msg 1
http://myaccount.queue.core.windows.net/myqueue/messages/messageid
<?xml version="1.0" encoding="utf-8"?>
?popreceipt=YzQ4Yzg1MDIGM0MDFiZDAwYzEw
<QueueMessagesList>
Msg 2
Msg 2
1

Web Role

<QueueMessage>
<MessageId>5974b586-0df3-4e2d-ad0c-18e3892bfca2</MessageId>
Msg 3
<InsertionTime>Mon, 22 Sep 2008 23:29:20 GMT</InsertionTime>
<ExpirationTime>Mon, 29 Sep 2008 23:29:20 GMT</ExpirationTime>
Msg 4
<PopReceipt>YzQ4Yzg1MDIGM0MDFiZDAwYzEw
</PopReceipt>
Worker Role
<TimeNextVisible>Tue, 23 Sep 2008 05:29:20GMT</TimeNextVisible>
<MessageText>PHRlc3Q+dG...dGVzdD4=</MessageText>
</QueueMessage>
Msg 2
</QueueMessagesList>
Queue

24

13.03.2009

What tables dont do

Not relational
No Referential Integrity
If theseNoareJoins
you,
Limited to
Queries
important
use:
No Group by

No Aggregations
No Transactions
No Transactions

Cheap

Very Scalable
Flexible
Durable

Data stored in Tables

What tables can do

A Table is a set of Entities (rows)


An Entity is a set of Properties (columns)

Entity has:

PartitionKey enables scalability


RowKey unique id within the partition
the only indexed property
Timestamp for optimistic concurrency
255 properties for your data
Max size of 1MB

25

13.03.2009

Vessel Position Reporting System SQL Server


VesselId

Time

Latitude

Longitude

Speed

xxx-xx1

10:15 14 Nov

01.23

53.24

xxx-xx1

10:05 14 Nov

04.45

54.32

xxx-xx1

09:55 14 Nov

02.32

52.34

xxx-xx2

10:15 14 Nov

01.23

51.23

10

To find last pos report for vessel in SQL:


select TOP(1) * from PosRpts
order by [Time] DESC
where VesselId = ???

Solving this the Azure way


PartitionKey
VesselId

Time

Latitude

Longitude

Speed

xxx-xx1

10:15 14 Nov

01.23

53.24

xxx-xx1

10:05 14 Nov

04.45

54.32

xxx-xx1

09:55 14 Nov

02.32

52.34

xxx-xx2

10:15 14 Nov

01.23

51.23

10

PartitionKey

26

13.03.2009

Stored in-order:
RowKey needs to
Just need to do a
be a string
top on the
partition PartitionKey
xxx-xx1
RowKey

Latitude

Longitude

Speed

2521756430999999999
10:15
14 Nov

01.23

53.24

2521756436999999999
10:05
14 Nov

04.45

54.32

09:55 14 Nov
Makes it
descending

02.32

52.34

2521756442999999999

100 nanoseconds!

(DateTime.MaxValue
time).Ticks.ToString
()
time.Ticks.ToString
()

Store Large Objects (up to 50 GB each)


Standard REST PUT/GET/DELETE Interface
http://<Account>.blob.core.windows.net/<Container>/<BlobName>

Account

Container

Blob

Block

IMG001.JPG

Pictures
IMG002.JPG

Account
Block AAAA

Movies

MOV1.AVI

Block AAAB

Block AAAC

27

13.03.2009

28

13.03.2009

29

13.03.2009

30

13.03.2009

31

13.03.2009

32

13.03.2009

33

13.03.2009

34

Buy

SAP

their instance
of SAP hosted
@ partner

Exchange

Exchange Online

Siebel

CRM Online

legacy HR system

Build

Control of Features
High

Low

Economy of Scale
Low

High

13.03.2009

to get access
to elastic
compute
and cheaper
storage

Clinical trial software


New molecule research

On Premise
High
Low

@Hoster

@Cloud

Control of SLA
Economy of Scale (EoS)

@Vendor
Low
High

35

Buy

SAP
Exchange

Exchange Online

Siebel

CRM Online

New molecule research

On Premise

Low

Big Challenge:
to
get access Corporate Boundaries
Crossing
to elastic
Business Processes
compute
Connectivity and Messaging
and cheaper
storageFederated Identity

Clinical trial software

High

their instance
of SAP hosted
@ partner

legacy HR system

Build

Control of Features
High

Low

Economy of Scale
Low

High

13.03.2009

@Hoster

Data Access
@Cloud
@Vendor

Control of SLA
Economy of Scale (EoS)

Low
High

Extension of .NET capabilities to the cloud


Leverage what you know to do things that are
otherwise pretty hard
3 Services today, more to follow

.NET Service Bus


connectivity and fan-out messaging necessary for
many integrations
.NET Workflow Service
reliably run workflows at scale
.NET Access Control Service
authorization based on federated identities

36

13.03.2009

Service Registry
Relay and direct connect connectivity
Publish/Subscribe
Integrated with Access Control services

[http|sb]://servicebus.windows.net/services/account/svc/
Service Registry Root

The service registry provides a


mapping from URIs to services

Root

servicebus.
windows.
net

services

account

svc

contoso

Multi-Tenant

37

13.03.2009

sb://servicebus.windows.net/services/user/service/endpoint

Relay
Outbound SSL-Secured TCP 828
Connection to Relay Rendezvous
Endpoint

Sender

One-Way Messages
through TCP Tunnel

Receiver

sb://servicebus.windows.net/services/user/service/endpoint

Relay
- Outbound SSL-Secured TCP 828
Connection to Relay
- Out-of-Band Protocol to
negotiate Direct Connection

Sender

Receiver
Upgrade to Direct
when possible

38

13.03.2009

Tooling
VS Designer

VS Debugger

Rehosted
Designer

Workflow
Activity
Library

WF Runtime
Hosts

IIS/WAS+
Dublin

Workflow
Service

your.exe
Direct

A reliable, scalable off-premises host for workflows

A portal at workflow.ex.azure.microsoft.com
New activities for the Azure Services
Platform
APIs that allow you to deploy, manage, and
run your workflows on the cloud
Enables you to orchestrate services

39

13.03.2009

Windows Azure Storage

Essential storage service in the cloud


Provides a core set of non-relational storage and retrieval
abstractions at massive scale

SQL Services

Premium database service in the cloud


Extends the rich capabilities of the SQL data platform to the cloud
at scale
Relational data processing over structured and unstructured
data
Integrate with key data platform capabilities e.g. Data
Analytics, Reporting, ETL

40

13.03.2009

Reporting
Data Mining

Reference Data

ETL

Data services tier of the Azure Services Platform


Built on SQL Server foundation
Broad data platform capabilities as a service

Friction-free provisioning, scaling


Significant investments in scale, HA, lights-out
operation and TCO

Unit of
geo-location
and billing
Tied to
DNS name
Collection
of Containers

41

13.03.2009

Tables
Stored Procedures
Triggers
Views
Indexes
Visual Studio Compatibility
ADO.Net Compatibility
ODBC Compatibility

Content

News
feeds
Upload
content

Data over
30 days

SQL
Services

Current
data

Large disk
arrays

42

13.03.2009

Current
solution

Download
data from
SQL
Services

Download
form app
Mobile
connection
to SQL
Services

Upload
e-form

SQL
Services

Publishing
application

Reports

Real-time data
Final results

Real-time data
Final results
Shared
Ownership

43

13.03.2009

creator

consumer
rich end-point

processor
creator

consumer

44

13.03.2009

45

13.03.2009

46

13.03.2009

The Web-Tier provides tremendous opportunities


Windows Azure An OS in the cloud for developing
and running applications in the Web-Tier
Azure Service Platform S+S building blocks

47

You might also like