1. Designing for scalability and high availability on Microsoft Azure
Microsoft Ignite 2016
9/22/2018 5:47 PM
BRK3207
Designing for scalability and high availability on Microsoft Azure
Adam Glick
Sr. Program Manager – Azure Resiliency
@MobileGlick
adamglick
© 2016 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

2. Session objectives and takeaways
Tech Ready 15
9/22/2018
Session objectives and takeaways
Session objective(s):
Design highly available applications on the Azure platform
Understand best practices and common pitfalls to avoid
Learn the inner workings of Azure and how to best work around hidden limitations
Note: This session focuses mostly on the design of Cloud-based apps using IaaS.
© 2012 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.

3. High availability What does HA mean to you?
What are the business requirements?
Different requirements mean a different architecture

4. Resiliency Resiliency Backup Disaster recovery High availability
Cost / Complexity
RPO/RTO
Backup
Disaster recovery
High availability
RTO >> 0
RTO > 0
RTO = 0
Best For:
Data deletion
Data corruption
Legal, governance & compliance
Protection for unplanned failures
Don’t want to re-architect for HA
Don’t want the cost of HA
Large scale failures
Mission critical apps
New apps
Localized failures

5. Mean time between failure (MTBF)
Metric used to identify reliability
Component failures
Cost of delivering cloud infrastructure
On-premises vs cloud
Hardware failure is inevitable

6. Designing for the cloud
No single point of failure
State stays at the edges of your app stack
Loosely couple your components
Build for scale
Process centrally, deliver locally
Automate
Trust and verify
Scale out, not up

7. Scaling vertical (up) vs horizontal (out)
Cost
VMSS and autoscale

8. Good candidates for Azure
Built versus bought software
Software designed for cloud
Applications designed around a single-state datastore

9. What can go wrong?
Compute
Networking
Storage
All 3

10. Compute

11. Availability Sets Governs fault and update domains
Tied to a role in your application
Required for 99.95% SLA from Microsoft
Provides fault domains and upgrade domains

12. Availability Sets (upgrade domain)
Fault Domain is single point of failure
Upgrade domain used for patching
ASM (2 FDs)
ARM (3 FDs) VM
Update Domain
Fault Domain
VM1
VM2 1
VM3 2
VM4 3
VM5 4 VM
Update Domain
Fault Domain
VM1
VM2 1
VM3 2
VM4 3
VM5 4
VM6 5

13. Deployment scale
Availability Set

14. Azure VMs
AVSet
FE Subnet
Resource Manager Template
Virtual Network

15. Azure VMs Front-end web servers Active Directory SQL Server cluster
Virtual Network
Availability Set
Subnet
Front-end web servers
Active Directory
SQL Server cluster
Resource Manager Template

16. Azure VMs East US West US Active Directory AD Repl. SQL AO Async
TechReady 23
9/22/2018 5:47 PM
Azure VMs
Virtual Network
Availability Set
Subnet
East US
West US
Active Directory
AD Repl.
SQL AO Async
Backup
Resource Manager Template
© 2016 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

17. Virtual Machine Scale Sets (VMSS)
A set of VMs which are identical (cattle)
Deployed with configuration management techniques
Can be auto-scaled
Based off a common base image
Portal / ARM template
Configuration management

18. Deployment scale - VMSS
Region 1

19. Deployment scale – multi-region

20. Network

21. Layer 4 vs layer 7 Load balancers
Application Gateway
3rd party solutions (NGINX, HAProxy, F5, Barracuda) L4
Availability Set L7 L7 L7

22. Demo Stages of availability

23. What do you do if a region fails?
a) Use a Load Balancer to shift to additional resources
b) Use Azure DNS to point to a paired region
c) Use Traffic Manager to change regions
d) Initiate your failover runbook in Azure Site Recovery
e) Complain on Twitter

24. How Traffic Manager handles endpoint loss
Update DNS
Zzzzzzz
You there?
Yup, all good.
GET X
No Response
200 OK
10s
20s
10s
20s
10s
20s
10s
20s
10s
20s
10s
Time

25. Traffic Manager DNS-based load balancing and failover
Health probe of endpoints to determine availability
Impact of Time To Live (TTL)
Multiple performance policies
Performance
Priority
Weighted

26. Storage

27. Storage Types of storage
Locally redundant (LRS) - 3 local copies strongly consistent
Zone redundant (ZRS) – 3 copies across 2 or 3 facilities
Geo-redundant (GRS) – 3 local copies strongly consistent, 3 remote copies eventually consistent
Read access geo-redundant (RA-GRS) – Same as GRS but with read-only access of eventually consistent data
Provides tables, blobs, queues, and files

28. Paired Regions Paired regions provide: Isolation Replication
Primary
Secondary
North Central US
South Central US
East US
West US
West US 2
West Central US
US East 2
Central US
North Europe
West Europe
South East Asia
East Asia
Germany Northeast
Germany Central
Primary
Secondary
East Asia
South East Asia
East China
North China
Japan East
Japan West
Brazil South
South Central US
Australia East
Australia Southeast
Canada Central
Canada East
UK South
UK West
Paired regions provide:
Isolation
Replication
Region order recovery
Sequential updates
Data residency

29. Storage – hidden point of failure?
Standard vs Premium
Disks live in a storage account
A storage account exists on a storage stamp
A storage stamp is a single point of failure
Storage outage can span fault domains !

30. Demo: Storage account information tool

31. Storage account information
TechReady 23
9/22/2018 5:47 PM
POST to (Yes, it’s an Azure Function):
Post Content:
{"storageUrl" : "storagetestblob1.blob.core.windows.net"}
Response (in JSON): {
"Region":"West US",
"Datacenter":"4",
"Stage":"Production",
"Type":"Standard",
"Stamp":"07",
"Version":"Primary" }
*This tool is not endorsed, supported, or in any way officially endorsed by Microsoft. Use at your own risk.
Send feedback to
© 2016 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

32. Common architectures

33. Consistency models Strong consistency Session consistency
Eventual consistency
Consistency Availability Partition Tolerant == CAP Theorem AP CP

34. Common single-region deployment- RDBMS
Storage
(LRS)
App
Availability
Set
Front End …
Database
Queue
Distributed Cache
Blobs

35. Multi-region deployment
Active/Passive
Only one region online with DR failover
Active/Active
Round robin (all regions accept traffic)
Active performance
Many regions, user is routed to nearest

36. Multi-region deployment for OLTP (active/passive)
Storage
(RA-GRS)
App
Availability
Set
Front End …
Database
Queue
Distributed Cache
Blobs
Storage
(RA-GRS)
App
Availability
Set
Front End …
Database
Queue
Blobs
Distributed Cache
Async
West US (Primary)
East US (Secondary)

37. Multi-region deployment (active/active)
Storage
(RA-GRS)
App
Availability
Set
Front End …
Database
Queue
Distributed Cache
Blobs
Storage
(RA-GRS)
App
Availability
Set
Front End …
Database
Queue
Distributed Cache
Blobs
Session
Consistency
West US (Primary)
East US (Secondary)

38. Demconvention.com Highly available Cross-regional
Thousands of requests per second
Under DDoS attack on Tuesday
Running on Azure

40. Using a CDN for availability
Protection from DDoS
Greater availability and scale for your origin
Other benefits:
Lower latency content delivery
Faster data delivery
Lower data usage
Origin obfuscation

41. Demo: Azure CDN

42. Fault Injection Chaos Monkey Tests for single point of failure
Automated architecture testing
Simulates failure

43. Session objectives and takeaways
Tech Ready 15
9/22/2018
Session objectives and takeaways
Session objective(s):
Design highly available applications on the Azure platform
Understand best practices and common pitfalls to avoid
Learn the inner workings of Azure and how to best work around hidden limitations
Note: This session focuses mostly on the design of Cloud-based apps using IaaS.
© 2012 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.

44. Resources Azure High Availability
High Availability Checklist –
Designing for High Availability –
Overview of HA and DR -

45. Related content
Backup born-in-the-cloud and hybrid applications with Operations Management Suite and Azure Backup 9/27 – 10:45AM-12:00PM
Review best practices for disaster recovery – from design to operations 9/27 – 2:15PM-3:30PM
Reinvent disaster recovery leveraging Microsoft Azure cloud infrastructure 9/28 – 2:15PM-3:30PM
Migrate and disaster recovery Azure workloads using Operations Management Suite 9/29 – 9:00AM-10:15AM
Run highly available solutions on Microsoft Azure 9/29 – 12:30PM-1:45PM

46. Free IT Pro resources To advance your career in cloud technology
Microsoft Ignite 2016
9/22/2018 5:47 PM
Free IT Pro resources To advance your career in cloud technology
Plan your career path
Microsoft IT Pro Career Center
Cloud role mapping
Expert advice on skills needed
Self-paced curriculum by cloud role
$300 Azure credits and extended trials
Pluralsight 3 month subscription (10 courses)
Phone support incident
Weekly short videos and insights from Microsoft’s leaders and engineers
Connect with community of peers and Microsoft experts
Get started with Azure
Microsoft IT Pro Cloud Essentials
Demos and how-to videos
Microsoft Mechanics
Connect with peers and experts
Microsoft Tech Community
© 2016 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

47. Free IT Pro resources To advance your career in cloud technology
Microsoft Ignite 2016
9/22/2018 5:47 PM
Free IT Pro resources To advance your career in cloud technology
Plan your career path
IT Pro Career Center
Get started with Azure
IT Pro Cloud Essentials
Demos and how-to videos
Microsoft Mechanics
Connect with peers and experts
Ask questions, get answers, exchange ideas
Azure Solutions
Get started with Azure Solutions today
Azure monthly webinar series
Join live or watch on-demand
© 2016 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

48. Please evaluate this session
9/22/2018 5:47 PM
Please evaluate this session
Your feedback is important to us!
From your PC or Tablet visit MyIgnite at
From your phone download and use the Ignite Mobile App by scanning the QR code above or visiting
© 2014 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

49. 9/22/2018 5:47 PM
Thank You!
© 2014 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

50. 9/22/2018 5:47 PM
© 2014 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

51. 9/22/2018
Q&A
If you have questions please proceed to the Q&A MICROPHONE located in your session room.
© 2015 Microsoft Corporation. All rights reserved. Microsoft, Windows, 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.

52. High availability checklist
Use Traffic Manager Avoid single VMs Use load balancers in front of web-facing VMs Put your stateless servers in Availability Sets Use VMSS for your stateless server scaling Use Premium Storage for your production VMs Use internal load balancers (or queues) between tiers Distribute your database Use caches Contact support before a high scale event Store static assets in Blob Storage Use a CDN in front of your static assets a a a a a a a a a a a