Enabling Data Residency and Data Protection in Azure Regions-2021
Enabling Data Residency and Data Protection in Azure Regions-2021
©2021 Microsoft Corporation. All rights reserved. This document is provided as is. Information and views expressed in this document, including
URL and other website references, may change without notice. You bear the risk of using it. This document does not provide you with any legal
rights to any intellectual property in any Microsoft product. You may copy and use this document for your internal reference purposes.
Contents
Introduction............................................................................................................................................................................................................. 5
Availability
Zone 1
Availability
Zone 3
Availability
Zone 2
Availability Zones
Availability Zones are unique physical locations within an Azure region. Each zone consists of
one or more datacenters equipped with independent power, cooling, and networking. Physical
separation of Availability Zones within a region protects applications and data from datacenter
failures.
An Availability Zone in an Azure region is a combination of a fault domain and an update domain.
Zone-redundant services replicate your applications and data across Availability Zones to protect
from single points of failure. This architecture also protects against unplanned downtime as well
as potential downtime from planned maintenance events. If one datacenter or one Availability
Zone fails, zone-redundant Azure services automatically replicate and continue in the other
Availability Zones without impacting the customer’s zonal applications. Moreover, if the Azure
platform is updating for faults or maintenance, the Azure platform recognizes this distribution
across update domains to make sure that VMs in different zones are not updated at the same
time.
With Availability Zones, Azure offers an industry-best 99.99% VM uptime service-level agreements.
Regions
A region is what the customer typically sees in the Azure portal or command line interface (CLI) as
a selectable scope for a deployment location. For example, customers can choose to deploy their
VMs into the region US West 2, which will create VMs in the physical location of the Azure US
West 2 datacenters. As illustrated in the graphic below, a region can consist of several Availability
Zones; for example, US West 2 consists of three Availability Zones. A region can also consist of
several datacenters even if the region does not have multiple Availability Zones.
Azure Region
Availability Zone 3
Japan East
Japan
Japan West
Korea Central
Korea
Korea South
Mexico Mexico Central (announced)
New Zealand New Zealand North (announced)
Norway East
Norway
Norway West
Poland Poland Central (announced)
Qatar Qatar Central (announced)
Spain Spain Central (announced)
The Azure regional concept allows customers to achieve two aspects of business continuity—
high availability and disaster recovery—while keeping the customer in control of data residency:
▪ High availability, for example, via Availability Zone SLAs with 99.99% VM uptime
▪ Disaster recovery via multiple zones in a region, a second region in the Azure geo, or
pairing to a region outside of an Azure geo
High availability
High availability refers to solutions that provide service availability, data availability, and
automatic recovery from failures that affect the service or data. service-level agreements (SLAs)
describe Microsoft commitments for uptime and connectivity. Availability Zones, as described
above, provide the highest uptime availability SLAs. In regions without Availability Zones,
Availability Sets (a logical grouping of VMs that provides for redundancy and availability) provide
99.95% uptime SLAs.
>See the full list of Azure SLAs in Microsoft Service-level agreements.3
High availability in regions with Availability Zones
VMs in an Availability Zone are synchronously replicated across the Availability Zone. If one zone
should fail, the VMs in the other zones will continue to run and Azure will load balance without
impacting the customer’s applications.
Managed Disks, which are like physical disks in an on-premises server but virtualized, deliver
consistent performance and high availability within Availability Zones as documented in the Disk
FAQs 4 and in Azure premium storage: design for high performance.5 Note that Managed Disks
also ensure that the placement of disks for VMs within an availability set (detailed below) honors
fault domain semantics, as documented in Availability options for Azure Virtual Machines.6
Managed Disks provide redundancy within an Availability Zone, with three replica instances
spread across storage stamps in the same datacenter. They also support designing for a recovery
point objective of zero hours within an Availability Zone for resiliency and high availability.
3https://aka.ms/Azure-SLA
4 https://aka.ms/Azure-Disc-Res
5 https://aka.ms/Azure-Prem-Stor I. Infrastructure of
6 https://aka.ms/Azure-VM-Avail Azure regions 8
Zone-redundant storage (ZRS) is a service that replicates Azure Storage data synchronously
across three Availability Zones within the same region. ZRS offers durability for Azure Storage
data objects of at least 99.9999999999% (twelve 9s) over a given year. With ZRS, data would still
be accessible for both read and write operations in the event a zone becomes unavailable. More
specifically, in the hypothetical situation where one Availability Zone fails, the Azure platform
would undertake networking updates, such as DNS repointing, to enable the other Availability
Zones to take on the storage workloads that are kept synchronous, allowing customers to design
for zero data loss.
>Refer to Zone-redundant storage 7 for further details.
High availability in regions without Availability Zones
In regions without Availability Zones, availability sets 8 provide 99.95% uptime SLAs. An
availability set is a logical grouping capability for isolating VM resources from each other when
they are deployed. Azure makes sure that the VMs you place within an availability set run across
multiple physical servers, compute racks, storage units, and network switches. If a hardware or
software failure occurs, only a subset of your VMs are impacted and your overall solution stays
operational.
Disaster recovery
All Azure regions are built for hyperscale production workloads. Azure services are designed
with redundancy to tolerate faults and minimize disruptions. Azure follows a rigorous testing and
production rollout process, which ensures that all technical components are in alignment, and
that customer solutions are not negatively impacted by deployment of new versions or processes.
Azure datacenters are designed to run 365 days a year, employing measures to protect
operations from physical intrusion, network failures, and power outages. Azure provides
hardware, network, local data redundancy, and Distributed Denial of Service (DDoS) protection.
Datacenters have dedicated uninterruptible power supplies (UPS) and emergency power
support, which includes onsite generators that provide backup power. Regular maintenance and
testing are conducted for both the UPS and generators, and operations teams have contractual
agreements with local vendors for emergency fuel delivery. Datacenters also have a dedicated
Facility Operations Center to monitor power systems, including critical electrical components.
Datacenters are required to test continued operation and resumption of critical datacenter
processes in the event of a disruption. Each critical service maintains and tests a disaster recovery
plan against each loss scenario to ensure restoration of service within recovery time and recovery
7 https://aka.ms/AZ-Stor-Redund point objectives. Any issues identified during testing are resolved, goals are set for continued
8 https://aka.ms/avail-sets improvement, and business continuity plans are updated accordingly.
9 https://aka.ms/AZ-VM-Migr I. Infrastructure of
10 https://aka.ms/AZ-VM-main Azure regions 9
Azure offers tools that customers can use to design highly available services, employing features
such as load balancing, Azure paired regions, Azure Backup, AzCopy, and Azure Storage
replication. Systems are proactively monitored to ensure service performance, and achieve
availability in accordance with financially backed service-level agreements.
>Customers can validate their disaster recovery strategies following the guidelines in Create and
customize recovery plans.11
11 https://aka.ms/AZ-Site-recovery
12 https://aka.ms/AZ-Site-
recover-over
13 https://aka.ms/redund-2nd- I. Infrastructure of
Azure regions 10
region
Not all Azure services automatically replicate data, nor do all Azure services automatically fail
back from a failed region to its pair. In such cases, customers must configure recovery and
replication. Explore which Azure high availability, disaster recovery, and backup capabilities to use
with your apps through the infographic below, Reliability with Azure.14
>You can find a listing of Azure regions pairing within the same geography in Business
continuity and disaster recovery: Azure Paired Regions.15
14 https://aka.ms/Reliability-
graphic
15 https://aka.ms/DR-paired- I. Infrastructure of
Azure regions 11
regions
At the time of this writing, Microsoft has announced the following Azure regions without a DR
region in the same geo.
▪ Austria East ▪ Mexico Central
▪ Chile Central ▪ New Zealand North
▪ Denmark East ▪ Poland Central
▪ Greece Central ▪ Qatar Central
▪ Indonesia Central ▪ Spain Central
▪ Israel Central ▪ Sweden Central
▪ Italy North ▪ Taiwan North
> See the availability by region of any Azure service at Products available by region.16
Azure disaster recovery services
Azure Site Recovery
Azure Site Recovery 17 helps ensure business continuity by keeping business apps and workloads
running during outages. It replicates workloads running on physical and virtual machines from
a primary region to a secondary region. When an outage occurs at the primary region, the
customer can trigger a failover to the secondary location and access applications from there. The
service is application-agnostic, enabling customers to build disaster recovery for any application
hosted on VMs to another zone, within the region or to another region. After the primary region
is healthy again, the customer can fail back to it.
Multi-VM consistency, a capability provided by Azure Site Recovery, creates a replication group
of all the machines. All of the machines in a replication group have shared crash-consistent and
app-consistent recovery points when failed over. Enabling multi-VM consistency can impact
workload performance as it is CPU intensive. The maximum number of VMs in a replication group
is sixteen.
▪ Crash-consistent recovery points capture data that was on the disk when the snapshot was
taken. This doesn’t include anything in memory, but it contains the equivalent of the on-disk
data that would be present if the VM crashed or the power was lost at the instant that the
snapshot was taken. A crash-consistent recovery point does not guarantee data consistency
for the operating system, nor for apps on the VM. Today, most apps can recover well from
crash-consistent recovery points. Azure Site Recovery creates crash-consistent recovery
points every five minutes by default.
▪ Application-consistent recovery points are created from app-consistent snapshots. An app-
consistent snapshot contains all the information in a crash-consistent snapshot, plus all the
data in memory and transactions in progress. App-consistent snapshots use the Volume
Shadow Copy Service (VSS): 1) when a snapshot is initiated, VSS performs a copy-on-write
(COW) operation on the volume; 2) before it performs the COW, VSS informs every app on
the machine that it needs to flush its memory-resident data to disk; and 3) VSS then allows
the backup/disaster recovery application (Azure Site Recovery) to read the snapshot data
and proceed.
App-consistent snapshots are more complex and take longer to complete than crash-
consistent snapshots and affect the performance of applications running on a VM enabled
for replication. By default, Azure Site Recovery takes an app-consistent snapshot every 4
hours, but it is possible to configure any value between 1 and 12 hours. Azure Site Recovery
keeps recovery points for 24 hours by default, but this can be configured to be a value
between 1 and 72 hours.
>Learn how to move Azure VMs to another region18 and enable zone-to-zone disaster
recovery19 for Azure VMs.
16 https://aka.ms/Products-
by-Region Azure Backup
17 https://aka.ms/AZ-Site-
recover-over The Azure Backup 20 service keeps your data safe and recoverable in case of a disaster. It allows
18 https://aka.ms/Azure- backups of entire Windows/Linux VMs (using backup extensions), or it can back up files, folders,
to-Azure and system state using the Microsoft Azure Recovery Services (MARS) agent.21 Backup is also
19 https://aka.ms/site-recovery-
available for Azure Files shares, SQL Server in Azure VMs, and SAP HANA databases running on
zone
20 https://aka.ms/azure-backup
Azure VMs.
I. Infrastructure of
21 https://aka.ms/MARS-agent Azure regions 12
Additional disaster recovery services
These include Azure DNS and Azure Traffic Manager,22 which enable customers to design a
resilient architecture that will survive the loss of the primary region, as well as AzCopy 23 to
schedule data backups to a storage account in a different region.
Because recovery time objectives (RTO) and recovery point objectives (RPO) are dependent on
customer architecture, Azure does not provide service-level agreements for RTO and RPO.
Latency considerations
Besides data residency, minimization of latency is also a key value proposition of local Azure
regions. Latency is a significant factor in the time it takes to transfer data between a primary and
secondary region. In disaster recovery, this is important because it determines whether and how
much data will be lost if the primary region fails. Latency between VMs affects the performance
of many applications. Bringing cloud services closer to customer workloads enables adoption of
latency-sensitive applications.
Accelerated networking
This is a feature whereby network traffic arrives at the VM’s network interface card (NIC), and the
NIC forwards network traffic directly to the VM, bypassing the host and the virtual switch.
>Get information about accelerated networking in Create a Windows VM with accelerated
22 https://aka.ms/Disaster-DNS networking using Azure PowerShell.27
23 https://aka.ms/Storage-
AZcopy Proximity placement groups
24 https://aka.ms/vnet-latency
25 https://aka.ms/expressroute-er
Proximity placement groups offer co-location of Azure VMs in the same datacenter. A proximity
26 https://aka.ms/about-fastpath
placement group is a logical grouping used to make sure that Azure compute resources are
27 https://aka.ms/vm-powershell
located physically close to each other, reducing latency.
28 https://aka.ms/proximity- >Read more in Introducing proximity placement groups.28 I. Infrastructure of
Azure regions 13
groups
Latency from Azure region to Azure region
Azure continuously monitors the latency of core areas of its network, using internal monitoring
tools as well as measurements collected by a third-party synthetic monitoring service
(ThousandEyes).
>Read more in Azure network round-trip latency statistics.29
In addition, a helpful latency test website is available to run a generic measurement of latency
from your location to any Azure region around the world.
29 https://aka.ms/AZ-latency
30 https://aka.ms/AZ-overview
31 https://aka.ms/Products-by- I. Infrastructure of
Azure regions 14
Region
II. Data residency for customer data
Data residency for regional services
As a customer, you retain all right, title, and interest in and to customer data—personal data and
other content—that you provide for storing and hosting in Azure services. Microsoft will not store
or process customer data outside the geography you specify, except for certain services and
scenarios that are discussed below. You are also in control of any additional geographies where
you decide to deploy your solutions or replicate your data. In addition, you and your users may
move, copy, or access your customer data from any location globally.
Most Azure services are deployed regionally and enable you to specify where your customer data
will be stored and processed. Examples of such regional services include VMs, storage, and SQL
Database. For a complete list, see Products available by region.32
For regional services, customers preselect the region in which the service will be deployed. (Note
that when you select the region in Azure, you automatically choose the geography that contains
that region.) Deployment location (and thereby data residency) can be defined by the region
variable in the Azure portal or via the command line interface.
INSTANCE DETAILS
Virtual machine name *
Availability
Zone 1
Availability
Zone 3
Availability
Zone 2
34 https://aka.ms/Azure-Data-Res
35 https://aka.ms/Azure-Data-Res
36https://aka.ms/LUIS-location
37https://aka.ms/AZ-serial-con
38https://aka.ms/Data-residency- II. Data residency for
customer data 16
AADMFA
▪ Azure Security Center stores a copy of security-related customer data, collected from or
associated with a customer resource (such as a VM or an Azure AD tenant), in the same geo
as that resource, except in those geos where Microsoft has yet to deploy Azure Security
Center, in which case a copy of such data will be stored in the United States. And where
Azure Security Center uses another Microsoft online service to process such data, it may
store it in accordance with the geolocation rules of that other online service.
>For more information, see Azure Security Center.39
▪ Services that provide global routing functions and do not themselves process
or store customer data. These services include Traffic Manager,40 which provides load
balancing between different regions, and Azure DNS,41 which provides domain name
services that route to different regions.
>For a complete list of non-regional services, see Products available by region42 and select
Non-regional for Region.
39https://aka.ms/AZ-sec-center
40https://aka.ms/AZ-traffic
41https://aka.ms/DNS-overview
42https://aka.ms/Products-by-
Region
43 https://aka.ms/AAD-Data-
Residency
44 https://aka.ms/AAD-Storage-
EU
45 https://aka.ms/
AZPolicySamples
46 https://aka.ms/ II. Data residency for
customer data 17
WhatIsAzurePolicy
Using Azure Blueprints to enforce compliance and
data residency
Azure Blueprints47 is a free service that supplies templates to create, deploy, and update fully
governed cloud environments to consistent standards, which helps customers comply with
regulatory requirements. It differs from Azure Resource Manager (ARM) and Azure Policy in
that Blueprints is a package that contains different types of artifacts—including ARM templates,
resource groups, policy assignments, and role assignments—all in one container, so you can
quickly and easily deploy all these components in a repeatable configuration. Blueprints help to
simplify large-scale Azure deployments by packaging policies in a single blueprint definition.
The Azure Blueprints service provides built-in blueprints mapped to key portions of common
standards such as HIPAA, FedRAMP, and PCI DSS. For example, the ISO 27001 Blueprint48 and
the PCI DSS Blueprint49 map a core set of policies for those respective standards to any Azure
environment. Blueprints can be deployed to multiple Azure subscriptions and managed from
a central location, and are scalable to support production implementations for large-scale
migrations.
You can use the built-in blueprints or create your own custom blueprints. Blueprints can be
created in the Azure portal or using the REST API with tools such as PowerShell. If the latter
method is used, you can create blueprint parameters to prevent conflicts when reusing certain
blueprints.
Blueprints can be used to help manage data residency for specific compliance needs by
specifying both allowed locations and allowed locations for resource groups—for example,
control mapping of the UK OFFICIAL and UK NHS blueprint samples.50 You can also use the
regulatory compliance dashboard for insight into your compliance posture based on how you're
meeting specific compliance requirements.
47 https://azure.microsoft.com/
services/blueprints/
48 https://aka.ms/ISO-27001-
Blueprint
49 https://aka.ms/PCI-DSS-
Blueprint
https://aka.ms/bp-parameters
50 https://aka.ms/UKOFFICIAL-
Blueprint II. Data residency for
customer data 18
https://aka.ms/comp-dashboard
III. Access to diagnostic, service-generated,
and support data
Microsoft relies on a set of definitions for types of data outlined in the Microsoft Online Services
Data Protection Addendum51 and Microsoft Online Services Terms.52
▪ Diagnostic data refers to data collected or obtained by Microsoft from software that is
locally installed by the customer for use with an Azure service. Diagnostic data may also be
referred to as telemetry. Diagnostic data does not include customer data, service-generated
data, or professional services data.
▪ Service-generated data refers to data generated or derived by Microsoft through the
operation of an Azure service. Service-generated data does not include customer data,
diagnostic data, or professional services data.
▪ Support data refers to all data, including all text, sound, video, image files, or software, that
are provided to Microsoft by or on behalf of an Azure customer to obtain technical support
for Azure services. Support data is a subset of professional services data.
For services where customer data is stored and processed in the cloud, service-generated,
diagnostic, and support data include application and server logs that are required to maintain
modern applications and platforms. These logs provide customers with the information they
need to operate and troubleshoot their workloads, and provide Microsoft with the information it
needs to operate, troubleshoot, and improve the platform.
Microsoft policies regarding data at rest in some cases do not apply to the three types of
data defined above. Microsoft uses such data for clearly defined scenarios, in line with GDPR
requirements and aligned to the best practices described in ISO/IEC 1994453 (the standard that
describes data flow, data categories, and data use in the cloud).
>See Who can access customer data and on what terms (page 23) for further information
regarding limitations on access to customer data, as well as diagnostic, service-generated, and
support data.
Diagnostic data
This includes data collected or obtained by Microsoft from software that is locally installed by the
customer for use with an Azure Service, such as log files, system-generated event logs, registry
keys, debug logs, server and database information, console screenshots, and basic network and
storage disk information.
For App Service-related issues, HTTP logs, detailed errors, KUDU trace, transform logs, FREB logs,
winsock logs, event logs, DAAS logs, and Webjob logs are collected to help with troubleshooting.
For Azure AD Connect-related issues, information about Active Directory objects (such as user
and device properties), your synchronization configuration, and related log files (such as Sign-In,
Audit, or synchronization logs) are collected to help with troubleshooting.
>Get a detailed list of diagnostic data that Microsoft collects in the following:
▪ Windows Server logs 54
▪ Azure PaaS VM logs 55
▪ Azure IaaS logs 56
51 https://aka.ms/MS-DPA ▪ Azure Service Fabric logs 57
52 https://aka.ms/Online-Services- ▪ StorSimple support package and device logs 58
Terms
53 https://www.iso.org/standard/ ▪ SQL Server on Azure VM logs 59
79573.html
54 https://aka.ms/WinServer-
▪ Azure Active Directory diagnostic logs 60
Logs-Diag
55 https://aka.ms/AZ-PaaS-logs
56 https://aka.ms/AZ-IaaS-logs
57 https://aka.ms/AZ-Serv-Fab-
logs
58 https://aka.ms/StorSimple-logs
59 https://aka.ms/SQL-VM-logs III. Access to diagnostic,
60https://aka.ms/AAD-dia-logs
service-generated, 19
and support data
Transparency and customer control of diagnostic data
Azure provides transparency and control functions for some of the most common types of
diagnostic data scenarios for Windows VMs, Linux VMs, and virtual networks.
>Get an overview of Azure Monitor agents.61
Windows VMs on Azure
Virtual machines and other compute resources require an agent to collect monitoring data to
measure the performance and availability of their guest operating system and workloads.
▪ The Azure Diagnostics extension62 collects monitoring data from the guest operating
system and workloads of Azure VMs and other compute resources.
▪ The Log Analytics agent63 collects monitoring data from the guest operating system and
workloads of VMs in Azure.
▪ The Dependency Agent64 collects discovered data about processes running on the VM and
external process dependencies.
Diagnostics Log Analytics Dependency
extension (WAD) agent agent
Azure Storage
Data sent to
Azure Monitor Metrics Azure Monitor Logs Azure Monitor Logs
Event Hub
61 https://aka.ms/AZ-Mon-Agents
Service-generated data
62 https://aka.ms/AZ-Mon-Diag Service health is an important aspect of operating Azure services, and so Microsoft collects
63 https://aka.ms/AZ-Mon-log schematized service-generated data to diagnose and perform root-cause analysis on incidents
64 https://aka.ms/AZ-Dep-Agent
for the platform. Every service uses this data to trigger self-healing processes, which reduces
65 https://aka.ms/Win-Sec-
the human intervention required. As an example, if the load on a specific component increases,
Baselines the platform assigns more resources to manage the load. Microsoft has integrated anonymized
66 https://aka.ms/Win-Sec- service-generated data in the Azure DevOps tools without accessing customer data.
Connect
67 https://aka.ms/AZ-Linux-Agent III. Access to diagnostic,
68 https://aka.ms/AZ-Mon-Agents
service-generated, 20
and support data
Customer access to service-generated data
▪ Azure Monitor.69
Customers can use Azure Monitor to manage the health of their workloads. This service
provides a 360-degree view of applications, infrastructure, and networks with advanced
analytics, dashboards, and visualization maps. Azure Monitor gives the customer a centralized
hub that helps to identify network glitches, CPU spikes, memory leaks in code, and other issues
before they impact the customer’s workload. Azure Monitor also offers various ways to notify
administrators in case of alerts, like Action rules, e-mail, SMS, or Logic Apps.
▪ Application Insights.70
Azure Monitor also includes Application Insights which provides an extensible Application
Performance Management (APM) service for web developers on multiple platforms. It can be
used to monitor web applications during runtime. It will automatically detect performance
anomalies, and it includes powerful analytics tools to help diagnose issues and understand
user interactions with applications.
Application Insights is designed to help continuously improve performance and usability by
sending service-generated data from the customer’s web applications to the Azure portal. It
works for apps on a wide variety of platforms, including .NET, Node.js, and J2EE, hosted on
premises or in the cloud. Collectors are designed to provide a schematized output of data,
limit the transmission of personal data as much as possible, and transmit data securely. A data
retention policy must be defined by the user. The customer can also use this data to build
high-availability workloads, which could detect an incident based on the data and perform
automated predefined actions to mitigate the incident.
▪ Azure Network Watcher.71
Customers can leverage Azure Network Watcher to monitor network traffic associated with
their IaaS workloads. This service provides tools to monitor, diagnose, view metrics, and
enable or disable logs for resources in an Azure virtual network. It is designed to monitor and
repair the network health of IaaS products, which include VMs, Virtual Networks, Application
Gateways, and Load Balancers. Note that Network Watcher is not intended for and will not
work for Platform as a Service (PaaS) monitoring or Web analytics.
Support data
Customers initiating a support request can give Microsoft engineers access to support data—
data that a customer provides to Microsoft to obtain technical support for Azure services.
Through the “Share diagnostic information” feature in the Azure portal, you can authorize a
Microsoft support engineer to remotely collect data related to the current support incident from
your Azure Virtual Machines or Azure Cloud to troubleshoot your issue. Support data can be
retained for up to 90 days.
>Get more information in Azure Support diagnostic information and memory dump
collection.72
Memory dump
When a customer VM crashes, customer data may be contained inside a memory dump file on
the VM. Customer data remains in the region where the VM is deployed, and does not leave the
Azure data residency boundary.
By default, Microsoft engineers do not have access to customer VMs and cannot review crash
dumps without the customer’s approval. Before investigating a VM crash dump, engineers must
gain explicit customer authorization to access customer crash dump data. Access is gated by
the Just-in-time (JIT) privileged access management system and Customer Lockbox so that all
but extraordinary actions, such as major outages and confidential law enforcement requests, are
logged and audited.
>Get more information on which services are covered by Customer Lockbox73 and on processes
69https://aka.ms/AZ-Monitor
for memory dump in Azure for Secure Worldwide Public Sector Cloud Adoption.74
70https://aka.ms/APP-Insights-
Over
71https://aka.ms/AZ-Netwatch
72 https://aka.ms/AZ-Legal-Diags
73 https://aka.ms/
III. Access to diagnostic,
msazurelockbox service-generated, 21
74 https://aka.ms/Azure-WWPS and support data
Enabling boot diagnostics
Customers can also choose to enable boot diagnostics, which captures logs, the serial console
output, and screenshots from the host running the VM. Enabling boot diagnostics also allows the
Azure platform to inspect the Operating System Virtual Hard Disk (OS VHD) for VM provisioning
errors, helping to provide deeper information on the root causes of failures. Access to the OS
VHD includes guest operating system information, system files on the OS VHD, and custom
scripts.
79 https://aka.ms/vm-dedicated
80 https://aka.ms/AZ-isolation-
choices
81 https://aka.ms/Azure-
subprocessors
82 https://aka.ms/MS-DPA
83 https://aka.ms/AZ-SCC
IV. How Microsoft protects
access to customer data 25
https://aka.ms/receive-notific
How Microsoft handles government requests
Microsoft has taken a firm public stand on protecting customer data from inappropriate
government access. Through clearly defined and well-established response policies and
processes, strong contractual commitments, and if need be, the courts, Microsoft defends your
data. We believe that all government requests for your data should be directed to you. We do not
give any government direct or unfettered access to customer data.
Microsoft is principled and transparent about how we respond to requests for data. Because
we believe that you have control over your own data, we will not disclose data to a government
except as you direct or where required by law. Microsoft scrutinizes all government demands to
ensure they are legally valid and appropriate.
If Microsoft receives a demand for a customer’s data, we will direct the requesting party to seek
the data directly from the customer. If compelled to disclose or give access to any customer’s
data, Microsoft will promptly notify the customer and provide a copy of the demand unless
legally prohibited from doing so.
>For data about enterprise customers who were impacted by law enforcement requests, see
Law enforcement requests (page 27).
Microsoft contractual commitments to our commercial and public sector customers include
Defending Your Data,84 which builds on our existing protections and demonstrates our
confidence in our ability to protect customer data against exposure to inappropriate disclosure:
▪ Microsoft will challenge every government request for commercial and public sector
customer data—from any government—where there is a lawful basis for doing so. We have
a proven track record of successfully using the courts to challenge government demands
that are inconsistent with the rule of law. We have more experience than any other company
taking the US government to court to challenge orders seeking access to an individual’s
data and to protect our ability to tell customers about those orders, even taking one case85
to the US Supreme Court. Our challenges have led to greater protections and transparency
for our customers worldwide, which include enabling us to disclose reports about the
number of US national security orders we receive.
▪ We stand behind the strength of our GDPR compliance and other data protection
safeguards. However, to provide added reassurance against liability for our commercial and
public sector customers, we will provide monetary compensation to these customer’s users
if we disclose their data in response to a government request in violation of the GDPR.
>For in-depth information about how Microsoft handles government requests, see About
our practices and your data.86
Key management
Key management can be performed by Azure or by the customer, and encryption can be
performed server-side or client-side.
▪ Server-side encryption: There are three server-side encryption models that offer
different key management characteristics from which you can choose according to your
organization’s requirements:
▪ Service-managed keys use Azure Key Vault to provide a combination of control and
convenience with low overhead. Azure resource providers perform the encryption and
decryption operations, and Microsoft manages the keys.
▪ Customer-managed keys give you control over the keys, including bring-your-own-key
(BYOK) support, or allow you to generate new ones. Azure resource providers perform the
encryption and decryption operations. The customer controls keys using Azure Key Vault.
>Learn more about configuring customer-managed keys with Azure Key Vault.99
▪ Customer-provided keys (CPK) enable you to store and manage keys in on-premises or
key stores other than Azure Key Vault.
▪ Client-side encryption: With client-side encryption, Microsoft does not have access to the
encryption keys and cannot decrypt the data. Customers encrypt data and upload the
data as an encrypted blob. The customer maintains complete control of the keys and keeps
keys on premises (or in other secure stores); keys are not available to Azure services. This
model is supported by Azure, but not by all Azure services.
>For more information, see Azure Encryption Overview.100
Key management: Server-side encryption
Service-managed keys
Azure Key Vault101 is a cloud-hosted service that provides centralized storage and management
of cryptographic keys and other secrets that are used in customers’ cloud applications. This
Azure service enables customers to safeguard cryptographic keys, certificates, and application
94 https://aka.ms/AZ-SQL-Encrypt
passwords, and helps protect secrets from accidental leakage.
95 https://aka.ms/AZ-SQL-Always Azure Key Vault uses specialized hardware security modules (HSMs) for maximum protection
96 https://aka.ms/AZ-DataLake and is designed in a way that allows customers to maintain control of keys and data. Usage
97 https://aka.ms/AZ-Cosmos-intro of customers’ stored keys can be monitored and audited in different ways, including Azure
98 https://aka.ms/AZ-Cosmo- logging and the import of these logs into Azure HDInsight. Customers can also incorporate this
Encrypt information into their existing security information and event management (SIEM) systems, which
99 https://aka.ms/AZ-CMK supports Microsoft customers in performing additional analysis, such as threat detection.
100 https://aka.ms/AZ-encryption-
overview V. How customers can protect data
101 https://aka.ms/AZ-keyvault-over from unauthorized access 29
Azure Key Vault allows segregation of secrets in multiple vaults. This helps reduce the chances of
accidental loss of security information by centralizing the storage of application secrets. Azure
Key Vault can handle requests and renewals of TLS certificates. It also provides features that
enable robust certificate lifecycle management. Note that Azure Key Vault is designed to support
application keys and secrets and is not intended to be a store for user passwords.
Access to a key vault is controlled through two separate interfaces: the management plane and
the data plane. Access controls for the management plane and data plane work independently.
Customers should use dedicated role definitions in Azure Active Directory to manage role-based
access. This approach implements an effective segregation of duties.
If you want an additional layer of security for encryption keys, you can add a key encryption key
(KEK) to your key vault. When a key encryption key is specified, Azure Disk Encryption uses that
key to wrap the data encryption key (DEK). The entity that has access to the KEK may be different
than the entity that requires the DEK. The DEK is cached and accessed by the resource provider
for efficient encryption as close to the data as possible. The KEK is under customer control in Key
Vault. By using the Azure Backup service,102 customers can back up and restore encrypted VMs
that use the KEK configuration.
Customer-managed keys
Azure Key Vault also provides a bring-your-own-key (BYOK) capability. Customers can generate
the keys on premises using an offline workstation equipped with an nCipher HSM, and then
transmit the keys securely to the Azure HSMs in the cloud. The nCipher software used for key
submission ensures that the keys are bound to this environment and can never be extracted out
of the HSMs. Customers who require additional functions such as enterprise key management
processes or hybrid cloud setups can use the CipherTrust Cloud Key Manager.
Storing the customer keys on premises eliminates the possibility for Azure to decrypt workloads,
though it also limits some Azure functionality, such as search.
>Get information about how Azure supports BYOK functionality in Import HSM-protected keys
to Key Vault.103
Customer-provided keys
These enable you to store and manage keys in on-premises or key stores other than Azure Key
Vault to meet corporate, contractual, and regulatory compliance requirements for data security.
Customer-provided keys (CPK) enable you to pass an encryption key as part of a read or write
operation to a storage service using blob APIs.104 Since the encryption key is defined at the
object level, you can have multiple encryption keys within a storage account. When you create
a blob with a customer-provided key, the storage service persists the SHA-256 hash of the
encryption key with the blob to validate future requests. When you retrieve an object, you must
provide the same encryption key as part of the request. For example, if a blob is created with
Put Blob105 using CPK, all subsequent write operations must provide the same encryption key. If
a different key is provided or if no key is provided in the request, the operation will fail with a 400
Bad Request. As the encryption key itself is provided in the request, a secure connection must be
established to transfer the key.
>Get more information about Customer Provided Keys with Azure Storage Service
Encryption.106
115 https://aka.ms/AZ-
expressroute-Enc
116 https://aka.ms/confidential-
compute
117 https://aka.ms/AZ-dcv2-series V. How customers can protect data
118 https://aka.ms/msazurelockbox from unauthorized access 32
VI. Data retention and deletion
Data retention and deletion policies and practices are important for protecting data. Privacy
compliance requires following fundamental principles for data retention. Microsoft follows strict
guidelines for retaining and deleting data.
In the Online Services Data Protection Addendum,119 Microsoft contractually commits to specific
processes when a customer terminates an online service or its subscription expires. This includes
deleting customer data from systems under Microsoft control.
Data retention
If a customer terminates a cloud subscription or it expires (with the exception of free trials),
Microsoft will store the customer’s data in a limited-function account for 90 days (the retention
period—a safety net, in effect) to enable customers to extract their data or renew their
subscription. During this period, Microsoft provides multiple notices so the customer will be
amply forewarned of the upcoming deletion of their data from Microsoft systems. Note, too, that
at all times during a subscription, customers can access, extract, and delete customer data stored
in our online services.
After this 90-day retention period, Microsoft will disable the account and delete the customer
data, including any cached or backup copies. For in-scope services, that deletion will occur
within 90 days after the end of the retention period. (In-scope services are defined in the Data
Processing Terms section of the Microsoft Online Services Terms.120
For Cognitive Services, a configuration or custom model that has been inactive may, for the
purposes of data retention and deletion, at Microsoft discretion, be treated as an online service
for which the customer’s subscription has expired. A configuration or custom model is inactive if
for 90 days: (1) no calls are made to it; (2) it has not been modified and does not have a current
key assigned to it and; (3) the customer has not signed in to it.
>For information in support of creating an exit plan for Microsoft cloud projects, see Exit
Planning for Microsoft Cloud Services.
Data deletion
Microsoft uses different types of data deletion techniques depending on the type of data object
that is being deleted—whole subscriptions, Azure Storage, Azure VMs, Azure SQL Database, or
Azure Active Directory.
▪ Subscriptions. As noted above, when a subscription is canceled or terminated, Microsoft
retains customer data for 90 days to permit the customer to extract its data. Microsoft
will then delete all customer data within another 90 days after the retention period (i.e.,
by day 180 after cancellation or termination). If a storage account is deleted within an
existing subscription (or when a subscription deletion has reached its timeout), the storage
account is not actually deleted for two weeks; this is to allow recovery from mistakes. When
a storage account is finally deleted, or when blob or table data is deleted outside the
context of a storage account deletion, the data is no longer available. To make storage data
unrecoverable faster, customers should delete tables and blobs individually before deleting
the storage account or canceling a subscription.
▪ Azure Storage. All disk writes are sequential in Azure Storage. This minimizes the number
of disk “seeks,” but requires updating the pointers to objects every time they are written.
(New versions of pointers are also written sequentially.) A side effect of this design is that
if there is a secret on disk, you can’t ensure it is gone by overwriting with other data. The
original data will remain on the disk and the new value will be written sequentially. Pointers
will be updated such that there is no longer any way to find the deleted value.
When the disk is full, the system has to write new logs onto disk space that has been freed
up by the deletion of old data. Instead of allocating log files directly from disk sectors,
log files are created in a file system running the New Technology File System (NTFS). A
background thread running on Azure Storage nodes frees up space by going through the
oldest log file, copying blocks that are still referenced from that oldest log file to the current
log file (and updating all pointers as it goes). It then deletes the oldest log file. Thus, there
119 https://aka.ms/MS-DPA
120 https://aka.ms/Online-Services- VI. Data retention and deletion 33
Terms
are two categories of free space on the disk: 1) space that NTFS knows is free, where it
allocates new log files from this pool and 2) space within those log files that Azure Storage
knows is free because there are no current pointers to it.
Customers can access only virtual disks and are never provided with access to the
underlying physical storage, so other customers and Microsoft personnel cannot read a
customer’s deleted data.
▪ Azure VMs are stored in Azure Storage as blobs, and the deletion rules apply as explained
in “Subscriptions” above. The virtualization mechanism is designed to ensure that those
spots on the disk cannot be read by another customer (or even by the same customer) until
data is written again. This mitigates the threat of data leakage. When a new virtual disk is
created for a VM, it will appear to the VM to be zeroed; however, the explicit zeroing of the
data buffers occurs when a portion of the virtual disk is read before it is written. If a VM
instance is reinitialized in place, it’s the same as if it had been moved to new hardware.
▪ Azure SQL Database. With Azure SQL Database, deleted data is marked for deletion. If
an entire database is deleted, it is the equivalent of deleting the database’s entire contents.
The SQL Database implementation is designed to ensure that user data is never leaked by
disallowing all access to the underlying storage except via the SQL Database API. That API
allows users to read, write, and delete data, but does not have a way to express the reading
of data that the user has not previously written.
▪ Azure Active Directory (AAD). When an administrator or the AAD service deletes a
user object, it is first moved into the recycle bin where the data remains intact. This gives a
customer the ability to easily recover user objects if they are accidentally deleted. After 30
days, the user object becomes a deleted object where the attribute data is removed from
AAD, except for the subset of unique identifier data that is required for replicating deletions
among Microsoft datacenters. At this point, no personal data remains. After another 30
days, the user object is removed from the AAD scale unit.
ISO/IEC 27018 Code of Practice for Protecting Personal Data in the Cloud
Azure complies with ISO/IEC 27018,127 which specifically targets controls that apply to processing
data in compliance with GDPR Article 9 requirements. At least once a year, Azure is audited for
its compliance with ISO/IEC 27018 by an accredited third-party certification body, providing
independent validation that applicable security controls are in place and operating effectively.
▪ Cyber Essentials Plus (UK) ▪ ISO 27001, 27018, 27017, and 27701
(PIMS) frameworks
▪ ENS (Spain)
▪ FedRAMP ▪ MTCS (Singapore)
>Shared Responsibility for Cloud Computing131 explains the roles and responsibilities that cloud
service providers and customers share in cloud computing.
126 https://aka.ms/MS-DPA
127 https://aka.ms/AZ-ISO27018
128 https://aka.ms/AZ-ISO27701
129 https://aka.ms/AZ-ISO27001
130 https://aka.ms/AZ-audit-
reports VII. Data residency and
131 https://aka.ms/AZ-Shared privacy compliance 36
Microsoft contractual commitments
When customers subscribe to an online service through a Microsoft Volume Licensing program,
we contractually guarantee our commitments in our standard contracts for commercial and
public sector customers. The terms that control how customers can use the service are defined
in the Microsoft Online Services Terms132 (OST) and the Online Services Data Protection
Addendum133 (DPA), both of which are available in 34 languages. Due to the frequency with
which Microsoft adds new services, the OST is updated monthly and the DPA is updated as
needed. Additional amendments exist to cover restricted industries, including financial services,
and will be available to customers where applicable. An archive contains older versions for
reference.
>Learn more about all the contractual commitments Microsoft makes in Licensing Terms,134 and
get the detailed terms that describe Microsoft commitments for Azure uptime and connectivity in
our service-level agreements.135
132 https://aka.ms/Online-
Services-Terms
133 https://aka.ms/MS-DPA
134 https://aka.ms/MS-product-
licensing
135 https://azure.microsoft.com/ VII. Data residency and
privacy compliance 37
support/legal/sla/
Appendix: Selected resources
Compliance guidance
Microsoft currently hosts datacenters in over 60 regions across multiple countries. Microsoft
provides compliance guidance for customers in a series of documents that address data
residency requirements generally, along with special emphasis on the financial services and
healthcare sectors.
▪ Navigating your way to the cloud in Europe:136 Belgium, Bulgaria, Croatia, Czech Republic,
Estonia, Finland, France, Germany, Italy, Ireland, Latvia, Lithuania, Luxembourg, Netherlands,
Norway, Poland, Slovenia, Spain, Sweden, Switzerland, United Kingdom
▪ Navigating your way to the cloud in Asia: A Guide for the Legal & Compliance
Professional:137 Australia, Bangladesh, Brunei, Hong Kong, India, Indonesia, Japan, Korea,
Malaysia, Nepal, New Zealand, Philippines, Singapore, Sri Lanka, Thailand, Vietnam
▪ Navigating your way to the cloud in the Middle East and Africa: A Guide for Legal and
Compliance Professionals:138 Angola, Jordan, Kenya, Mauritius, Morocco, Nigeria, Rwanda,
South Africa, UAE
136 https://aka.ms/TrustedCloud-
Europe
137 https://aka.ms/TrustedCloud-
APAC
138 https://aka.ms/TrustedCloud-
MEA
139 https://www.microsoft.com/
trust-center
140 https://aka.ms/AzureWWPS
141 https://aka.ms/AZ-Data- Appendix:
Selected Resources 38
Privacy