Choosing Your Windows Azure Platform Strategy

Marcus Tillett
T:@drmarcustillett

www.dotnetsolutions.co.uk

 What is the Windows Azure Platform
 Scenarios for Azure
 Strategies for Moving to Azure
 Challenges and Opportunities

 Currently 6 data centres in 3 regions
 Europe, US, Asia
 Purpose-built data centre to
accommodate containers at large
scale
 Cost $500 million, 100,000 square foot
facility (10 football fields)
 40 foot shipping containers can
house as many as 2,500 servers
 Density of 10 times amount of
compute in equivalent space in
traditional data centre

 First Mega Data Centre built outside the
US, on-line on July 2009
 Cost $500 million, 303+sq ft facility and
growing
 EU data centre Best Practice Award Winner
 Stand-alone Server Pods
▪ Environmentally Sustainable
▪ Free Air-Cooling via Air-Side Economization
 50% less energy use vs. traditional facilities
 1 % water use vs. traditional facilities
 1.25 PUE
 5.4 Mega Watts today, 22.2 MW for future
use
 ISO 27001:2005 Accreditation
 SAS 70 Type I and II Attestations

 Server in someone else’s data centre
 Commodity hardware
 Design for failure not against
 On demand compute and storage capability

Private Infrastructure Platform
(On-Premise) (as a Service) (as a Service)

You manage
Applications Applications Applications

You manage
Runtimes Runtimes Runtimes

Security & Integration Security & Integration Security & Integration

Managed by vendor
Databases Databases Databases
You manage

Servers Servers Servers

Managed by vendor
Virtualization Virtualization Virtualization

Server HW Server HW Server HW

Storage Storage Storage

Networking Networking Networking

Application Services

Frameworks “Dublin” “Velocity”

Security Access Control “Geneva”

Project
Connectivity Service Bus “Sydney”

SQL Azure
Data Data Sync

Compute
Content
Storage Table Storage Blob Storage Queue Drive Delivery
Network

 Development, service hosting, & management environment
 .NET, Java PHP, Python, Ruby, native code (C/C++, Win32, etc.)
 ASP.NET providers, FastCGI, memcached, MySQL, Tomcat
 Full-trust – supports standard languages and APIs
 Secure certificate store
 Management API’s, and logging and diagnostics systems
 Multiple roles – Web, Worker
 Multiple VM sizes
 1.6 GHz CPU x64, 1.75GB RAM, 100Mbps network, 250GB volatile storage
 Small (1X), Medium (2X), Large (4X), X-Large (8X)
 In-place rolling upgrades, organized by upgrade domains
 Walk each upgrade domain one at a time
 No remote desktop management or access to physical resources

Compute

 Rich data abstractions – tables, blobs, queues, drives, CDN
 Capacity (100TB), throughput (100MB/sec), transactions (1K req/sec)
 High accessibility
 Supports geo-location
 Language & platform agnostic REST APIs
 URL: http://<account>.<store>.core.windows.net
 Client libraries for .NET, Java, PHP, etc.
 High durability – data is replicated 3 times within a cluster
 High scalability – data is automatically partitioned and load
balanced across servers

Storage

 Highly available, scalable, and consistent distributed relational
database; geo-replication and geo-location of data
 Relational database, provided as a service
 Highly symmetrical development and tooling experience (use TDS protocol and
T-SQL)
 Highly scaled out, on commodity hardware
 Built on the SQL Server technology foundation
 Editions: Web (1GB), Business (10GB)

Data

 Securely connect applications
 Over the internet
 Across any network topology
 Across organizational boundaries
 Primary application patterns
 Eventing – notify applications and/or
devices
 Service Remoting – securely project on-
premises services out to the cloud
 Tunneling – app-to-app communication
Connectivity Service Bus with NAT/Firewall traversal

Compute

Average
Usage
Time

 Manage explosive growth
 Limit capital expenditure

Load Balancer

Stateless Web Roles

Stateless Worker Roles

Table / Blob Relational
State Tier Queue
Storage DB

Compute

Time

 Services with micro seasonality trends
 Peaks due to periodic increased demand

Peak Usage

Normal
Usage

10 60
Servers Servers

Compute

Time

 Unexpected/unplanned peak in demand
 Sudden spike impacts performance
 Can’t over provision for extreme cases

 http://blog.maartenballiauw.be/post/2010/06/02/Runn
ing-on-Windows-Azure-e28093-ChronoRace.aspx

Compute

Inactivity
Average Period Usage
Time
 On & off workloads (e.g. batch job)
 Short or unpredictable lifetime applications

 An application that archives data
 Serving data at massive scale
 Pure storage solution
 SQL Azure – TDS
 Windows Azure Storage – REST

Table Storage Blob Storage Queue Drive

 Connect on-premises apps with the cloud
 Create cached-mode clients
 Built on Microsoft Sync Framework and
ADO.NET Sync Services

Sync Sync /
Cache

Identity
Service
Trust

Connectivity
Service
3
Authenticate

Connect Open
4 2 1
5 Authenticate

Relayed
Connection

Client 6 Service

Direct
Connection
(optional)

 Greenfield
 Start-up or application
 Migration
 Full applications or part
 Service oriented architectures
 Distributed applications
 Loosely couple application architectures
 Few architectural compromises
 Extension
 Provide extra capacity
 Split architecture

 .NET 3.5 SP1 or .NET 4.0
 IIS 7.0
 64 bit
 ASP.NET web application (not web site)
 Core SQL Server components
 Abstract logging, configuration etc
 Avoid using O/S features
 Avoid legacy code and applications
(especially any that require installs)

 Stateless
 Asynchronous
 Distributed
 Design for failure
 Hardware
 Network (e.g. blips)

 Compute
 Web and worker role often biggest cost
 2 required for SLA
 Consider on/off for tasks
 Group tasks together under single role
 Compression
 Reduce data storage cost
 Reduce bandwidth cost
 Improve performance and latency for data read/write
 Data
 Consolidate database
 Migrate to table/blob storage from SQL Azure

 Data
 Data portability issues
 Migrating data
 How much does data sovereignty matter
 What data would you never put in the cloud
 Security
 Does data in the cloud pose a security risk
compared to on-premise
 Shared hardware impacts security
 Data/hardware destruction policies

 Legislative
 Data protection laws
 Taxation treatment
 Jurisdiction challenges
 Compliance requirements
 Availability
 Sending data over the Internet is slow
 The Internet is not always available
 Reliability in service providers’ hands

 Not a “magic bullet” for current scalability
issues
 No in-built SMTP/FTP support
 No out of the box backup or DR (yet)

 Server patching managed by platform
 Deployment and environment managed by
platform
 Monitoring managed by platform
 Familiar development environment
 Symmetrical platform
 Massively scalable
 Highly flexible model
 Pay per usage
 3 fold durable data storage
 Interoperable

 Further reading
 http://www.azure.com/
 http://blog.smarx.com/
 Signup for Azure
 http://www.microsoft.com/windowsazure/offers/
(Credit card required)
 Ensure that you complete delete a deployment –
otherwise you will be charged!
 Join your local user group
 UK (http://ukazurenet.com)
 Buy a book

 http://tinyurl.com/azurebook

 “Thinking of... Delivering
Solutions on the Windows
Azure Platform?”

 ISBN: 0956155634

C Infinity Zone
16:45, Friday, 11th June

Choosing Your Windows Azure Platform Strategy

More Related Content

What's hot (17)

Viewers also liked (7)

Similar to Choosing Your Windows Azure Platform Strategy (20)

Recently uploaded (20)

Choosing Your Windows Azure Platform Strategy